{
    "paper_id": "N19-1015",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T14:01:14.796499Z"
    },
    "title": "Topic-Guided Variational Autoencoders for Text Generation",
    "authors": [
        {
            "first": "Wenlin",
            "middle": [],
            "last": "Wang",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": "wenlin.wang@duke.edu"
        },
        {
            "first": "Zhe",
            "middle": [],
            "last": "Gan",
            "suffix": "",
            "affiliation": {},
            "email": "zhe.gan@microsoft.com"
        },
        {
            "first": "Hongteng",
            "middle": [],
            "last": "Xu",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": "hongteng.xu@infiniaml.com"
        },
        {
            "first": "Ruiyi",
            "middle": [],
            "last": "Zhang",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": "ruiyi.zhang@duke.edu"
        },
        {
            "first": "Guoyin",
            "middle": [],
            "last": "Wang",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Dinghan",
            "middle": [],
            "last": "Shen",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": ""
        },
        {
            "first": "Changyou",
            "middle": [],
            "last": "Chen",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "University at Buffalo",
                "location": {}
            },
            "email": "changyou@buffalo.edu"
        },
        {
            "first": "Lawrence",
            "middle": [],
            "last": "Carin",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Duke University",
                "location": {}
            },
            "email": "lcarin@duke.edu"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "We propose a topic-guided variational autoencoder (TGVAE) model for text generation. Distinct from existing variational autoencoder (VAE) based approaches, which assume a simple Gaussian prior for the latent code, our model specifies the prior as a Gaussian mixture model (GMM) parametrized by a neural topic module. Each mixture component corresponds to a latent topic, which provides guidance to generate sentences under the topic. The neural topic module and the VAE-based neural sequence module in our model are learned jointly. In particular, a sequence of invertible Householder transformations is applied to endow the approximate posterior of the latent code with high flexibility during model inference. Experimental results show that our TGVAE outperforms alternative approaches on both unconditional and conditional text generation, which can generate semantically-meaningful sentences with various topics.",
    "pdf_parse": {
        "paper_id": "N19-1015",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "We propose a topic-guided variational autoencoder (TGVAE) model for text generation. Distinct from existing variational autoencoder (VAE) based approaches, which assume a simple Gaussian prior for the latent code, our model specifies the prior as a Gaussian mixture model (GMM) parametrized by a neural topic module. Each mixture component corresponds to a latent topic, which provides guidance to generate sentences under the topic. The neural topic module and the VAE-based neural sequence module in our model are learned jointly. In particular, a sequence of invertible Householder transformations is applied to endow the approximate posterior of the latent code with high flexibility during model inference. Experimental results show that our TGVAE outperforms alternative approaches on both unconditional and conditional text generation, which can generate semantically-meaningful sentences with various topics.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "Text generation plays an important role in various natural language processing (NLP) applications, such as machine translation (Cho et al., 2014; Sutskever et al., 2014) , dialogue generation , and text summarization (Nallapati et al., 2016; Rush et al., 2015) . As a competitive solution to this task, the variational autoencoder (VAE) (Kingma and Welling, 2013; has been widely used in text-generation systems (Bowman et al., 2015; Serban et al., 2017) . In particular, VAE defines a generative model that propagates latent codes drawn from a simple prior through a decoder to manifest data samples. The generative model is further augmented with an inference network, that feeds observed data samples through an encoder to yield a distribution on the corresponding latent codes.",
                "cite_spans": [
                    {
                        "start": 127,
                        "end": 145,
                        "text": "(Cho et al., 2014;",
                        "ref_id": "BIBREF8"
                    },
                    {
                        "start": 146,
                        "end": 169,
                        "text": "Sutskever et al., 2014)",
                        "ref_id": "BIBREF44"
                    },
                    {
                        "start": 217,
                        "end": 241,
                        "text": "(Nallapati et al., 2016;",
                        "ref_id": "BIBREF31"
                    },
                    {
                        "start": 242,
                        "end": 260,
                        "text": "Rush et al., 2015)",
                        "ref_id": null
                    },
                    {
                        "start": 337,
                        "end": 363,
                        "text": "(Kingma and Welling, 2013;",
                        "ref_id": "BIBREF19"
                    },
                    {
                        "start": 412,
                        "end": 433,
                        "text": "(Bowman et al., 2015;",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 434,
                        "end": 454,
                        "text": "Serban et al., 2017)",
                        "ref_id": "BIBREF38"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "Compared with other potential methods, e.g., those based on generative adversarial networks (GANs) Guo et al., 2017; Zhang et al., 2017b , VAE is of particular interest when one desires not only text generation, but also the capacity to infer meaningful latent codes from text. Ideally, semanticallymeaningful latent codes can provide high-level guidance while generating sentences. For example, when generating text, the vocabulary could potentially be narrowed down if the input latent code corresponds to a certain topic (e.g., the word \"military\" is unlikely to appear in a sports-related document).",
                "cite_spans": [
                    {
                        "start": 99,
                        "end": 116,
                        "text": "Guo et al., 2017;",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 117,
                        "end": 136,
                        "text": "Zhang et al., 2017b",
                        "ref_id": "BIBREF56"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "However, in practice this desirable property is not fully achieved by existing VAE-based text generative models, because of the following two challenges. First, the sentences in documents may associate with different semantic information (e.g., topic, sentiment, etc.) while the latent codes of existing VAE-based text generative models often employ a simple Gaussian prior, which cannot indicate the semantic structure among sentences and may reduce the generative power of the decoder. Although some variants of VAE try to impose some structure on the latent codes (Jiang et al., 2016; Dilokthanakul et al., 2016) , they are often designed with pre-defined parameter settings without incorporating semantic meanings into the latent codes, which may lead to over-regularization (Dilokthanakul et al., 2016) .",
                "cite_spans": [
                    {
                        "start": 567,
                        "end": 587,
                        "text": "(Jiang et al., 2016;",
                        "ref_id": "BIBREF16"
                    },
                    {
                        "start": 588,
                        "end": 615,
                        "text": "Dilokthanakul et al., 2016)",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 779,
                        "end": 807,
                        "text": "(Dilokthanakul et al., 2016)",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The second issue associated with VAE-based text generation is \"posterior collapse,\" first identified in Bowman et al. (2015) . With a strong auto-regressive decoder network (e.g., LSTM), the model tends to ignore the information from the latent code and merely depends on previous generated tokens for prediction. Several strategies are proposed to mitigate this problem, including making the decoder network less auto-regressive (i.e., \u00b5 log 2 \u2713 < l a t e x i t s h a 1 _ b a s e 6 4 = \" 2 U n l 4 N N J R P X a u u g M u c 4 E V m Z 8 m A 8 = \" > A A A B + n i c b V B P S 8 M w H E 3 n v z n / 1 X n 0 E h y C p 9 G K o N 6 G X j x O s D p Y y 0 j T d A t L 0 5 L 8 K o 6 y r + L F g 4 p X P 4 k 3 v 4 3 p 1 o N u P g h 5 v P f 7 k Z c X Z o J r c J x v q 7 a y u r a + U d 9 s b G 3 v 7 O 7 Z + 8 1 7 n e a K M o + m I l W 9 k G g m u G Q e c B C s l y l G k l C w h 3 B 8 X f o P j 0 x p n s o 7 m G Q s S M h Q 8 p h T A k Y a 2 E 0 / T E W k J 4 m 5 s A 8 j B m R g t 5 y 2 M w N e J m 5 F W q h C d 2 B / + V F K 8 4 R J o I J o 3 X e d D I K C K O B U s G n D z z X L C B 2 T I e s b K k n C d F D M s k / x s V E i H K f K H A l 4 p v 7 e K E i i y 3 h m M i E w 0 o t e K f 7 n 9 X O I L 4 K C y y w H J u n 8 o T g X G F J c F o E j r h g F M T G E U M V N V k x H R B E K p q 6 G K c F d / P I y 8 U 7 b l 2 3 3 9 q z V u a r a q K N D d I R O k I v O U Q f d o C 7 y E E V P 6 B m 9 o j d r a r 1 Y 7 9 b H f L R m V T s H 6 A + s z x + W Q Z R Y < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 2 U n l 4 N N J R P X a u u g M u c 4 E V m Z 8 m A 8 = \" > A A A B + n i c b V B P S 8 M w H E 3 n v z n / 1 X n 0 E h y C p 9 G K o N 6 G X j x O s D p Y y 0 j T d A t L 0 5 L 8 K o 6 y r + L F g 4 p X P 4 k 3 v 4 3 p 1 o N u P g h 5 v P f 7 k Z c X Z o J r c J x v q 7 a y u r a + U d 9 s b G 3 v 7 O 7 Z + 8 1 7 n e a K M o + m I l W 9 k G g m u G Q e c B C s l y l G k l C w h 3 B 8 X f o P j 0 x p n s o 7 m G Q s S M h Q 8 p h T A k Y a 2 E 0 / T E W k J 4 m 5 s A 8 j B m R g t 5 y 2 M w N e J m 5 F W q h C d 2 B / + V F K 8 4 R J o I J o 3 X e d D I K C K O B U s G n D z z X L C B 2 T I e s b K k n C d F D M s k / x s V E i H K f K H A l 4 p v 7 e K E i i y 3 h m M i E w 0 o t e K f 7 n 9 X O I L 4 K C y y w H J u n 8 o T g X G F J c F o E j r h g F M T G E U M V N V k x H R B E K p q 6 G K c F d / P I y 8 U 7 b l 2 3 3 9 q z V u a r a q K N D d I R O k I v O U Q f d o C 7 y E E V P 6 B m 9 o j d r a r 1 Y 7 9 b H f L R m V T s H 6 A + s z x + W Q Z R Y < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 2 U n l 4 N N J R P X a u u g M u c 4 E V m Z 8 m A 8 = \" > A A A B + n i c b V B P S 8 M w H E 3 n v z n / 1 X n 0 E h y C p 9 G K o N 6 G X j x O s D p Y y 0 j T d A t L 0 5 L 8 K o 6 y r + L F g 4 p X P 4 k 3 v 4 3 p 1 o N u P g h 5 v P f 7 k Z c X Z o J r c J x v q 7 a y u r a + U d 9 s b G 3 v 7 O 7 Z + 8 1 7 n e a K M o + m I l W 9 k G g m u G Q e c B C s l y l G k l C w h 3 B 8 X f o P j 0 x p n s o 7 m G Q s S M h Q 8 p h T A k Y a 2 E 0 / T E W k J 4 m 5 s A 8 j B m R g t 5 y 2 M w N e J m 5 F W q h C d 2 B / + V F K 8 4 R J o I J o 3 X e d D I K C K O B U s G n D z z X L C B 2 T I e s b K k n C d F D M s k / x s V E i H K f K H A l 4 p v 7 e K E i i y 3 h m M i E w 0 o t e K f 7 n 9 X O I L 4 K C y y w H J u n 8 o T g X G F J c F o E j r h g F M T G E U M V N V k x H R B E K p q 6 G K c F d / P I y 8 U 7 b l 2 3 3 9 q z V u a r a q K N D d I R O k I v O U Q f d o C 7 y E E V P 6 B m 9 o j d r a r 1 Y 7 9 b H f L R m V T s H 6 A + s z x + W Q Z R Y < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 2 U n l 4 N N J R P X a u u g M u c 4 E V m Z 8 m A 8 = \" > A A A B + n i c b V B P S 8 M w H E 3 n v z n / 1 X n 0 E h y C p 9 G K o N 6 G X j x O s D p Y y 0 j T d A t L 0 5 L 8 K o 6 y r + L F g 4 p X P 4 k 3 v 4 3 p 1 o N u P g h 5 v P f 7 k Z c X Z o J r c J x v q 7 a y u r a + U d 9 s b G 3 v 7 O 7 Z + 8 1 7 n e a K M o + m I l W 9 k G g m u G Q e c B C s l y l G k l C w h 3 B 8 X f o P j 0 x p n s o 7 m G Q s S M h Q 8 p h T A k Y a 2 E 0 / T E W k J 4 m 5 s A 8 j B m R g t 5 y 2 M w N e J m 5 F W q h C d 2 B / + V F K 8 4 R J o I J o 3 X e d D I K C K O B U s G n D z z X L C B 2 T I e s b K k n C d F D M s k / x s V E i H K f K H A l 4 p v 7 e K E i i y 3 h m M i E w 0 o t e K f 7 n 9 X O I L 4 K C y y w H J u n 8 o T g X G F J c F o E j r h g F M T G E U M V N V k x H R B E K p q 6 G K c F d / P I y 8 U 7 b l 2 3 3 9 q z V u a r a q K N D d I R O k I v O U Q f d o C 7 y E E V P 6 B m 9 o j d r a r 1 Y 7 9 b H f L R m V T s H 6 A + s z x + W Q Z R Y < / l a t e x i t > Neural Topic Model (NTM) d t < l a t e x i t s h a 1 _ b a s e 6 4 = \" u o 7 N n e 9 4 D t v o M i f d 7 m u o d O j 9 Q X M = \" > A A A B 8 3 i c b V B N S w M x E J 2 t X 7 V + V T 1 6 C R b B U 9 k V Q b 0 V v X i s 4 G q h X U o 2 m 2 1 D s 8 m a Z A t l 6 e / w 4 k H F q 3 / G m / / G b L s H b R 0 I e b w 3 w 7 x 5 Y c q Z N q 7 7 7 V R W V t f W N 6 q b t a 3 t n d 2 9 + v 7 B g 5 a Z I t Q n k k v V C b G m n A n q G 2 Y 4 7 a S K 4 i T k 9 D E c 3 R T 6 4 5 g q z a S 4 N 5 O U B g k e C B Y z g o 2 l g l 4 o e a Q n i f 2 Q 6 d c b b t O d F V o G X g k a U F a 7 X / / q R Z J k C R W G c K x 1 1 3 N T E + R Y G U Y 4 n d Z 6 m a Y p J i M 8 o F 0 L B U 6 o D v K Z 6 S k 6 s U y E Y q n s E w b N 2 N 8 T O U 5 0 4 c x 2 J t g M 9 a J W k P 9 p 3 c z E l 0 H O R J o Z K s h 8 U Z z Z + y Q q E k A R U 5 Q Y P r E A E 8 W s V 0 S G W G F i b E 4 1 G 4 K 3 e P I y 8 M + a V 0 3 v 7 r z R u i 7 T q M I R H M M p e H A B L b i F N v h A 4 A m e 4 R X e n L H z 4 r w 7 H / P W i l P O H M K f c j 5 / A E S M k f c = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" u o 7 N n e 9 4 D t v o M i f d 7 m u o d O j 9 Q X M = \" > A A A B 8 3 i c b V B N S w M x E J 2 t X 7 V + V T 1 6 C R b B U 9 k V Q b 0 V v X i s 4 G q h X U o 2 m 2 1 D s 8 m a Z A t l 6 e / w 4 k H F q 3 / G m / / G b L s H b R 0 I e b w 3 w 7 x 5 Y c q Z N q 7 7 7 V R W V t f W N 6 q b t a 3 t n d 2 9 + v 7 B g 5 a Z I t Q n k k v V C b G m n A n q G 2 Y 4 7 a S K 4 i T k 9 D E c 3 R T 6 4 5 g q z a S 4 N 5 O U B g k e C B Y z g o 2 l g l 4 o e a Q n i f 2 Q 6 d c b b t O d F V o G X g k a U F a 7 X / / q R Z J k C R W G c K x 1 1 3 N T E + R Y G U Y 4 n d Z 6 m a Y p J i M 8 o F 0 L B U 6 o D v K Z 6 S k 6 s U y E Y q n s E w b N 2 N 8 T O U 5 0 4 c x 2 J t g M 9 a J W k P 9 p 3 c z E l 0 H O R J o Z K s h 8 U Z z Z + y Q q E k A R U 5 Q Y P r E A E 8 W s V 0 S G W G F i b E 4 1 G 4 K 3 e P I y 8 M + a V 0 3 v 7 r z R u i 7 T q M I R H M M p e H A B L b i F N v h A 4 A m e 4 R X e n L H z 4 r w 7 H / P W i l P O H M K f c j 5 / A E S M k f c = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" u o 7 N n e 9 4 D t v o M i f d 7 m u o d O j 9 Q X M = \" > A A A B 8 3 i c b V B N S w M x E J 2 t X 7 V + V T 1 6 C R b B U 9 k V Q b 0 V v X i s 4 G q h X U o 2 m 2 1 D s 8 m a Z A t l 6 e / w 4 k H F q 3 / G m / / G b L s H b R 0 I e b w 3 w 7 x 5 Y c q Z N q 7 7 7 V R W V t f W N 6 q b t a 3 t n d 2 9 + v 7 B g 5 a Z I t Q n k k v V C b G m n A n q G 2 Y 4 7 a S K 4 i T k 9 D E c 3 R T 6 4 5 g q z a S 4 N 5 O U B g k e C B Y z g o 2 l g l 4 o e a Q n i f 2 Q 6 d c b b t O d F V o G X g k a U F a 7 X / / q R Z J k C R W G c K x 1 1 3 N T E + R Y G U Y 4 n d Z 6 m a Y p J i M 8 o F 0 L B U 6 o D v K Z 6 S k 6 s U y E Y q n s E w b N 2 N 8 T O U 5 0 4 c x 2 J t g M 9 a J W k P 9 p 3 c z E l 0 H O R J o Z K s h 8 U Z z Z + y Q q E k A R U 5 Q Y P r E A E 8 W s V 0 S G W G F i b E 4 1 G 4 K 3 e P I y 8 M + a V 0 3 v 7 r z R u i 7 T q M I R H M M p e H A B L b i F N v h A 4 A m e 4 R X e n L H z 4 r w 7 H / P W i l P O H M K f c j 5 / A E S M k f c = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" u o 7 N n e 9 4 D t v o M i f d 7 m u o d O j 9 Q X M = \" > A A A B 8 3 i c b V B N S w M x E J 2 t X 7 V + V T 1 6 C R b B U 9 k V Q b 0 V v X i s 4 G q h X U o 2 m 2 1 D s 8 m a Z A t l 6 e / w 4 k H F q 3 / G m / / G b L s H b R 0 I e b w 3 w 7 x 5 Y c q Z N q 7 7 7 V R W V t f W N 6 q b t a 3 t n d 2 9 + v 7 B g 5 a Z I t Q n k k v V C b G m n A n q G 2 Y 4 7 a S K 4 i T k 9 D E c 3 R T 6 4 5 g q z a S 4 N 5 O U B g k e C B Y z g o 2 l g l 4 o e a Q n i f 2 Q 6 d c b b t O d F V o G X g k a U F a 7 X / / q R Z J k C R W G c K x 1 1 3 N T E + R Y G U Y 4 n d Z 6 m a Y p J i M 8 o F 0 L B U 6 o D v K Z 6 S k 6 s U y E Y q n s E w b N 2 N 8 T O U 5 0 4 c x 2 J t g M 9 a J W k P 9 p 3 c z E l 0 H O R J o Z K s h 8 U Z z Z + y Q q E k A R U 5 Q Y P r E A E 8 W s V 0 S G W G F i b E 4 1 G 4 K 3 e P I y 8 M + a V 0 3 v 7 r z R u i 7 T q M I R H M M p e H A B L b i F N v h A 4 A m e 4 R X e n L H z 4 r w 7 H / P W i l P O H M K f c j 5 / A E S M k f c = < / l a t e x i t >",
                "cite_spans": [
                    {
                        "start": 104,
                        "end": 124,
                        "text": "Bowman et al. (2015)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" 5 j P 2 h O 2 B T r 5 e g X K w s 6 6 h W I X X U p M = \" > A A A B + X i c b V C 9 T s M w G P x S / k r 5 S 2 F k s a i Q m K o E I Q F b B Q t j k Q i t 1 E a V 4 z i t V c e J b A d U h T 4 K C w M g V t 6 E j b f B a T N A y 0 m W T 3 f f J 5 8 v S D l T 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 1 / f v V Z J J Q j 2 S 8 E R 2 A 6 w o Z 4 J 6 m m l O u 6 m k O A 4 4 7 Q T j 6 8 L v P F C p W C L u 9 C S l f o y H g k W M Y G 2 k g V 3 v B w k P 1 S Q 2 F + o H V O O B 3 X C a z g x o m b g l a U C J 9 s D + 6 o c J y W I q N O F Y q Z 7 r p N r P s d S M c D q t 9 T N F U 0 z G e E h 7 h g o c U + X n s + h T d G y U E E W J N E d o N F N / b + Q 4 V k U 6 M x l j P V K L X i H + 5 / U y H V 3 4 O R N p p q k g 8 4 e i j C O d o K I H F D J J i e Y T Q z C R z G R F Z I Q l J t q 0 V T M l u I t f X i b e a f O y 6 d 6 e N V p X Z R t V O I Q j O A E X z q E F N 9 A G D w g 8 w j O 8 w p v 1 Z L 1 Y 7 9 b H f L R i l T s H 8 A f W 5 w + x s Z P U < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 5 j P 2 h O 2 B T r 5 e g X K w s 6 6 h W I X X U p M = \" > A A A B + X i c b V C 9 T s M w G P x S / k r 5 S 2 F k s a i Q m K o E I Q F b B Q t j k Q i t 1 E a V 4 z i t V c e J b A d U h T 4 K C w M g V t 6 E j b f B a T N A y 0 m W T 3 f f J 5 8 v S D l T 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 1 / f v V Z J J Q j 2 S 8 E R 2 A 6 w o Z 4 J 6 m m l O u 6 m k O A 4 4 7 Q T j 6 8 L v P F C p W C L u 9 C S l f o y H g k W M Y G 2 k g V 3 v B w k P 1 S Q 2 F + o H V O O B 3 X C a z g x o m b g l a U C J 9 s D + 6 o c J y W I q N O F Y q Z 7 r p N r P s d S M c D q t 9 T N F U 0 z G e E h 7 h g o c U + X n s + h T d G y U E E W J N E d o N F N / b + Q 4 V k U 6 M x l j P V K L X i H + 5 / U y H V 3 4 O R N p p q k g 8 4 e i j C O d o K I H F D J J i e Y T Q z C R z G R F Z I Q l J t q 0 V T M l u I t f X i b e a f O y 6 d 6 e N V p X Z R t V O I Q j O A E X z q E F N 9 A G D w g 8 w j O 8 w p v 1 Z L 1 Y 7 9 b H f L R i l T s H 8 A f W 5 w + x s Z P U < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 5 j P 2 h O 2 B T r 5 e g X K w s 6 6 h W I X X U p M = \" > A A A B + X i c b V C 9 T s M w G P x S / k r 5 S 2 F k s a i Q m K o E I Q F b B Q t j k Q i t 1 E a V 4 z i t V c e J b A d U h T 4 K C w M g V t 6 E j b f B a T N A y 0 m W T 3 f f J 5 8 v S D l T 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 1 / f v V Z J J Q j 2 S 8 E R 2 A 6 w o Z 4 J 6 m m l O u 6 m k O A 4 4 7 Q T j 6 8 L v P F C p W C L u 9 C S l f o y H g k W M Y G 2 k g V 3 v B w k P 1 S Q 2 F + o H V O O B 3 X C a z g x o m b g l a U C J 9 s D + 6 o c J y W I q N O F Y q Z 7 r p N r P s d S M c D q t 9 T N F U 0 z G e E h 7 h g o c U + X n s + h T d G y U E E W J N E d o N F N / b + Q 4 V k U 6 M x l j P V K L X i H + 5 / U y H V 3 4 O R N p p q k g 8 4 e i j C O d o K I H F D J J i e Y T Q z C R z G R F Z I Q l J t q 0 V T M l u I t f X i b e a f O y 6 d 6 e N V p X Z R t V O I Q j O A E X z q E F N 9 A G D w g 8 w j O 8 w p v 1 Z L 1 Y 7 9 b H f L R i l T s H 8 A f W 5 w + x s Z P U < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 5 j P 2 h O 2 B T r 5 e g X K w s 6 6 h W I X X U p M = \" > A A A B + X i c b V C 9 T s M w G P x S / k r 5 S 2 F k s a i Q m K o E I Q F b B Q t j k Q i t 1 E a V 4 z i t V c e J b A d U h T 4 K C w M g V t 6 E j b f B a T N A y 0 m W T 3 f f J 5 8 v S D l T 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 1 / f v V Z J J Q j 2 S 8 E R 2 A 6 w o Z 4 J 6 m m l O u 6 m k O A 4 4 7 Q T j 6 8 L v P F C p W C L u 9 C S l f o y H g k W M Y G 2 k g V 3 v B w k P 1 S Q 2 F + o H V O O B 3 X C a z g x o m b g l a U C J 9 s D + 6 o c J y W I q N O F Y q Z 7 r p N r P s d S M c D q t 9 T N F U 0 z G e E h 7 h g o c U + X n s + h T d G y U E E W J N E d o N F N / b + Q 4 V k U 6 M x l j P V K L X i H + 5 / U y H V 3 4 O R N p p q k g 8 4 e i j C O d o K I H F D J J i e Y T Q z C R z G R F Z I Q l J t q 0 V T M l u I t f X i b e a f O y 6 d 6 e N V p X Z R t V O I Q j O A E X z q E F N 9 A G D w g 8 w j O 8 w p v 1 Z L 1 Y 7 9 b H f L R i l T s H 8 A f W 5 w + x s Z P U < / l a t e x i t > y 0 < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > y 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" W e B U F x b q o F 6 j D 1 k 2 X + 8 U v e v Y q 3 0 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s J + 3 S z S b s b o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T A X X x n W / n d L K 6 t r 6 R n m z s r W 9 s 7 t X 3 T 9 4 1 E m m G P o s E Y l q h 1 S j 4 B J 9 w 4 3 A d q q Q x q H A V j i 6 m f q t J 1 S a J / L B j F M M Y j q Q P O K M G i v d j 3 t e r 1 p z 6 + 4 M Z J l 4 B a l B g W a v + t X t J y y L U R o m q N Y d z 0 1 N k F N l O B M 4 q X Q z j S l l I z r A j q W S x q i D f H b q h J x Y p U + i R N m S h s z U 3 x M 5 j b U e x 6 H t j K k Z 6 k V v K v 7 n d T I T X Q Y 5 l 2 l m U L L 5 o i g T x C R k + j f p c 4 X M i L E l l C l u b y V s S B V l x q Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 g j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H r 6 j X U = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W e B U F x b q o F 6 j D 1 k 2 X + 8 U v e v Y q 3 0 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s J + 3 S z S b s b o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T A X X x n W / n d L K 6 t r 6 R n m z s r W 9 s 7 t X 3 T 9 4 1 E m m G P o s E Y l q h 1 S j 4 B J 9 w 4 3 A d q q Q x q H A V j i 6 m f q t J 1 S a J / L B j F M M Y j q Q P O K M G i v d j 3 t e r 1 p z 6 + 4 M Z J l 4 B a l B g W a v + t X t J y y L U R o m q N Y d z 0 1 N k F N l O B M 4 q X Q z j S l l I z r A j q W S x q i D f H b q h J x Y p U + i R N m S h s z U 3 x M 5 j b U e x 6 H t j K k Z 6 k V v K v 7 n d T I T X Q Y 5 l 2 l m U L L 5 o i g T x C R k + j f p c 4 X M i L E l l C l u b y V s S B V l x q Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 g j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H r 6 j X U = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W e B U F x b q o F 6 j D 1 k 2 X + 8 U v e v Y q 3 0 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s J + 3 S z S b s b o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T A X X x n W / n d L K 6 t r 6 R n m z s r W 9 s 7 t X 3 T 9 4 1 E m m G P o s E Y l q h 1 S j 4 B J 9 w 4 3 A d q q Q x q H A V j i 6 m f q t J 1 S a J / L B j F M M Y j q Q P O K M G i v d j 3 t e r 1 p z 6 + 4 M Z J l 4 B a l B g W a v + t X t J y y L U R o m q N Y d z 0 1 N k F N l O B M 4 q X Q z j S l l I z r A j q W S x q i D f H b q h J x Y p U + i R N m S h s z U 3 x M 5 j b U e x 6 H t j K k Z 6 k V v K v 7 n d T I T X Q Y 5 l 2 l m U L L 5 o i g T x C R k + j f p c 4 X M i L E l l C l u b y V s S B V l x q Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 g j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H r 6 j X U = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W e B U F x b q o F 6 j D 1 k 2 X + 8 U v e v Y q 3 0 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s J + 3 S z S b s b o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T A X X x n W / n d L K 6 t r 6 R n m z s r W 9 s 7 t X 3 T 9 4 1 E m m G P o s E Y l q h 1 S j 4 B J 9 w 4 3 A d q q Q x q H A V j i 6 m f q t J 1 S a J / L B j F M M Y j q Q P O K M G i v d j 3 t e r 1 p z 6 + 4 M Z J l 4 B a l B g W a v + t X t J y y L U R o m q N Y d z 0 1 N k F N l O B M 4 q X Q z j S l l I z r A j q W S x q i D f H b q h J x Y p U + i R N m S h s z U 3 x M 5 j b U e x 6 H t j K k Z 6 k V v K v 7 n d T I T X Q Y 5 l 2 l m U L L 5 o i g T x C R k + j f p c 4 X M i L E l l C l u b y V s S B V l x q Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 g j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H r 6 j X U = < / l a t e x i t > y M < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t >",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "y M 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > y M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" h b g B F f + t P e K 6 W p k Q R V F C Q T 6 n 2 x Y = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E C s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 G L x 5 U v P q H v P l v 3 L Y 5 a O u D g c d 7 M 8 z M C 1 P B t X H d b 6 e 0 s r q 2 v l H e r G x t 7 + z u V f c P H n W S K Y Y + S 0 S i 2 i H V K L h E 3 3 A j s J 0 q p H E o s B W O b q Z + 6 w m V 5 o l 8 M O M U g 5 g O J I 8 4 o 8 Z K / r i X 3 0 1 6 1 Z p b d 2 c g y 8 Q r S A 0 K N H v V r 2 4 / Y V m M 0 j B B t e 5 4 b m q C n C r D m c B J p Z t p T C k b 0 Q F 2 L J U 0 R h 3 k s 2 M n 5 M Q q f R I l y p Y 0 Z K b + n s h p r P U 4 D m 1 n T M 1 Q L 3 p T 8 T + v k 5 n o M s i 5 T D O D k s 0 X R Z k g J i H T z 0 m f K 2 R G j C 2 h T H F 7 K 2 F D q i g z N p + K D c F b f H m Z + G f 1 q 7 p 3 f 1 5 r X B d p l O E I j u E U P L i A B t x C E 3 x g w O E Z X u H N k c 6 L 8 + 5 8 z F t L T j F z C H / g f P 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A a R e O n Q = = < / l a t e x i t > y 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > y 2 < l a t e x i t s h a 1 _ b a s e 6 4 = \" O c",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "+ V 6 Q Q q c S K 5 Y o 3 A h E K Y a f a R F n k = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 m K o N 6 K X j x W M L b Q h r L Z T t q l m 0 3 Y 3 Q g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m A q u j e t + O 6 W 1 9 Y 3 N r f J 2 Z W d 3 b / + g e n j 0 q J N M M f R Z I h L V C a l G w S X 6 h h u B n V Q h j U O B 7 X B 8 O / P b T 6 g 0 T + S D m a Q Y x H Q o e c Q Z N V b y J / 2 8 M e 1 X a 2 7 d n Y O s E q 8 g N S j Q 6 l e / e o O E Z T F K w w T V u u u 5 q Q l y q g x n A q e V X q Y x p W x M h 9 i 1 V N I Y d Z D P j 5 2 S M 6 s M S J Q o W 9 K Q u f p 7 I q e x 1 p M 4 t J 0 x N S O 9 7 M 3 E / 7 x u Z q K r I O c y z Q x K t l g U Z Y K Y h M w + J w O u k B k x s Y Q y x e 2 t h I 2 o o s z Y f C o 2 B G / 5 5 V X i N + r X d e / + o t a 8 K d I o w w m c w j l 4 c A l N u I M W + M C A w z O 8 w p s j n R f n 3 f l Y t J a c Y u Y Y / s D 5 / A F A K 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "6 C < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" O c",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "+ V 6 Q Q q c S K 5 Y o 3 A h E K Y a f a R F n k = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 m K o N 6 K X j x W M L b Q h r L Z T t q l m 0 3 Y 3 Q g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m A q u j e t + O 6 W 1 9 Y 3 N r f J 2 Z W d 3 b / + g e n j 0 q J N M M f R Z I h L V C a l G w S X 6 h h u B n V Q h j U O B 7 X B 8 O / P b T 6 g 0 T + S D m a Q Y x H Q o e c Q Z N V b y J / 2 8 M e 1 X a 2 7 d n Y O s E q 8 g N S j Q 6 l e / e o O E Z T F K w w T V u u u 5 q Q l y q g x n A q e V X q Y x p W x M h 9 i 1 V N I Y d Z D P j 5 2 S M 6 s M S J Q o W 9 K Q u f p 7 I q e x 1 p M 4 t J 0 x N S O 9 7 M 3 E / 7 x u Z q K r I O c y z Q x K t l g U Z Y K Y h M w + J w O u k B k x s Y Q y x e 2 t h I 2 o o s z Y f C o 2 B G / 5 5 V X i N + r X d e / + o t a 8 K d I o w w m c w j l 4 c A l N u I M W + M C A w z O 8 w p s j n R f n 3 f l Y t J a c Y u Y Y / s D 5 / A F A K 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "6 C < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" O c",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "+ V 6 Q Q q c S K 5 Y o 3 A h E K Y a f a R F n k = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 m K o N 6 K X j x W M L b Q h r L Z T t q l m 0 3 Y 3 Q g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m A q u j e t + O 6 W 1 9 Y 3 N r f J 2 Z W d 3 b / + g e n j 0 q J N M M f R Z I h L V C a l G w S X 6 h h u B n V Q h j U O B 7 X B 8 O / P b T 6 g 0 T + S D m a Q Y x H Q o e c Q Z N V b y J / 2 8 M e 1 X a 2 7 d n Y O s E q 8 g N S j Q 6 l e / e o O E Z T F K w w T V u u u 5 q Q l y q g x n A q e V X q Y x p W x M h 9 i 1 V N I Y d Z D P j 5 2 S M 6 s M S J Q o W 9 K Q u f p 7 I q e x 1 p M 4 t J 0 x N S O 9 7 M 3 E / 7 x u Z q K r I O c y z Q x K t l g U Z Y K Y h M w + J w O u k B k x s Y Q y x e 2 t h I 2 o o s z Y f C o 2 B G / 5 5 V X i N + r X d e / + o t a 8 K d I o w w m c w j l 4 c A l N u I M W + M C A w z O 8 w p s j n R f n 3 f l Y t J a c Y u Y Y / s D 5 / A F A K 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "6 C < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" O c",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "+ V 6 Q Q q c S K 5 Y o 3 A h E K Y a f a R F n k = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 m K o N 6 K X j x W M L b Q h r L Z T t q l m 0 3 Y 3 Q g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m A q u j e t + O 6 W 1 9 Y 3 N r f J 2 Z W d 3 b / + g e n j 0 q J N M M f R Z I h L V C a l G w S X 6 h h u B n V Q h j U O B 7 X B 8 O / P b T 6 g 0 T + S D m a Q Y x H Q o e c Q Z N V b y J / 2 8 M e 1 X a 2 7 d n Y O s E q 8 g N S j Q 6 l e / e o O E Z T F K w w T V u u u 5 q Q l y q g x n A q e V X q Y x p W x M h 9 i 1 V N I Y d Z D P j 5 2 S M 6 s M S J Q o W 9 K Q u f p 7 I q e x 1 p M 4 t J 0 x N S O 9 7 M 3 E / 7 x u Z q K r I O c y z Q x K t l g U Z Y K Y h M w + J w O u k B k x s Y Q y x e 2 t h I 2 o o s z Y f C o 2 B G / 5 5 V X i N + r X d e / + o t a 8 K d I o w w m c w j l 4 c A l N u I M W + M C A w z O 8 w p s j n R f n 3 f l Y t J a c Y u Y Y / s D 5 / A F A K 4 6 C < / l a t e x i t > y 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > y 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > GRU GRU ( T )",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" 9 D h B j R K F / p a 5 5 R c h 5 C 9 5 t z A g p 6 c = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H V 4 l v I K M L J Y V E h l q R K E B G w V L I x F a m i l J o o c x 2 2 t 2 n F k O 0 h V 1 F 9 g 4 V d Y G A C x M r L x N z h t B t p y J M t H 5 9 y r e + + J U k a V d p w f a 2 V 1 b X 1 j s 7 J V 3 d 7 Z 3 d u 3 D w 4 f l M g k J h 4 W T M h u h B R h N C G e p p q R b i o J 4 h E j n W h 0 W / i d R y I V F U l b j 1 M S c D R I a J 9 i p I 0 U 2 n U / E i x W Y 2 4 + 6 C s 6 4 A j O a x H R K G y f h X b N a T h T w G X i l q Q G S r R C + 9 u P B c 4 4 S T R m S K m e 6 6 Q 6 y J H U F D M y q f q Z I i n C I z Q g P U M T x I k K 8 u l F E 3 h q l B j 2 h T Q v 0 X C q / u 3 I E V f F g q a S I z 1 U i 1 4 h / u f 1 M t 2 / C n K a p J k m C Z 4 N 6 m c M a g G L e G B M J c G a j Q 1 B W F K z K 8 R D J B H W J s S q C c F d P H m Z e O e N 6 4 Z 7 f 1 F r 3 p R p V M A x O A F 1 4 I J L 0 A R 3 o A U 8 g M E T e A F v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "4 N 1 6 t l 6 t D + t z V r p i l T 1 H Y A 7 W 1 y 9 + V Z z + < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 9 D h B j R K F / p a 5 5 R c h 5 C 9 5 t z A g p 6 c = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H V 4 l v I K M L J Y V E h l q R K E B G w V L I x F a m i l J o o c x 2 2 t 2 n F k O 0 h V 1 F 9 g 4 V d Y G A C x M r L x N z h t B t p y J M t H 5 9 y r e + + J U k a V d p w f a 2 V 1 b X 1 j s 7 J V 3 d 7 Z 3 d u 3 D w 4 f l M g k J h 4 W T M h u h B R h N C G e p p q R b i o J 4 h E j n W h 0 W / i d R y I V F U l b j 1 M S c D R I a J 9 i p I 0 U 2 n U / E i x W Y 2 4 + 6 C s 6 4 A j O a x H R K G y f h X b N a T h T w G X i l q Q G S r R C + 9 u P B c 4 4 S T R m S K m e 6 6 Q 6 y J H U F D M y q f q Z I i n C I z Q g P U M T x I k K 8 u l F E 3 h q l B j 2 h T Q v 0 X C q / u 3 I E V f F g q a S I z 1 U i 1 4 h / u f 1 M t 2 / C n K a p J k m C Z 4 N 6 m c M a g G L e G B M J c G a j Q 1 B W F K z K 8 R D J B H W J s S q C c F d P H m Z e O e N 6 4 Z 7 f 1 F r 3 p R p V M A x O A F 1 4 I J L 0 A R 3 o A U 8 g M E T e A F v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "4 N 1 6 t l 6 t D + t z V r p i l T 1 H Y A 7 W 1 y 9 + V Z z + < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 9 D h B j R K F / p a 5 5 R c h 5 C 9 5 t z A g p 6 c = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H V 4 l v I K M L J Y V E h l q R K E B G w V L I x F a m i l J o o c x 2 2 t 2 n F k O 0 h V 1 F 9 g 4 V d Y G A C x M r L x N z h t B t p y J M t H 5 9 y r e + + J U k a V d p w f a 2 V 1 b X 1 j s 7 J V 3 d 7 Z 3 d u 3 D w 4 f l M g k J h 4 W T M h u h B R h N C G e p p q R b i o J 4 h E j n W h 0 W / i d R y I V F U l b j 1 M S c D R I a J 9 i p I 0 U 2 n U / E i x W Y 2 4 + 6 C s 6 4 A j O a x H R K G y f h X b N a T h T w G X i l q Q G S r R C + 9 u P B c 4 4 S T R m S K m e 6 6 Q 6 y J H U F D M y q f q Z I i n C I z Q g P U M T x I k K 8 u l F E 3 h q l B j 2 h T Q v 0 X C q / u 3 I E V f F g q a S I z 1 U i 1 4 h / u f 1 M t 2 / C n K a p J k m C Z 4 N 6 m c M a g G L e G B M J c G a j Q 1 B W F K z K 8 R D J B H W J s S q C c F d P H m Z e O e N 6 4 Z 7 f 1 F r 3 p R p V M A x O A F 1 4 I J L 0 A R 3 o A U 8 g M E T e A F v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "4 N 1 6 t l 6 t D + t z V r p i l T 1 H Y A 7 W 1 y 9 + V Z z + < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 9 D h B j R K F / p a 5 5 R c h 5 C 9 5 t z A g p 6 c = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H V 4 l v I K M L J Y V E h l q R K E B G w V L I x F a m i l J o o c x 2 2 t 2 n F k O 0 h V 1 F 9 g 4 V d Y G A C x M r L x N z h t B t p y J M t H 5 9 y r e + + J U k a V d p w f a 2 V 1 b X 1 j s 7 J V 3 d 7 Z 3 d u 3 D w 4 f l M g k J h 4 W T M h u h B R h N C G e p p q R b i o J 4 h E j n W h 0 W / i d R y I V F U l b j 1 M S c D R I a J 9 i p I 0 U 2 n U / E i x W Y 2 4 + 6 C s 6 4 A j O a x H R K G y f h X b N a T h T w G X i l q Q G S r R C + 9 u P B c 4 4 S T R m S K m e 6 6 Q 6 y J H U F D M y q f q Z I i n C I z Q g P U M T x I k K 8 u l F E 3 h q l B j 2 h T Q v 0 X C q / u 3 I E V f F g q a S I z 1 U i 1 4 h / u f 1 M t 2 / C n K a p J k m C Z 4 N 6 m c M a g G L e G B M J c G a j Q 1 B W F K z K 8 R D J B H W J s S q C c F d P H m Z e O e N 6 4 Z 7 f 1 F r 3 p R p V M A x O A F 1 4 I J L 0 A R 3 o A U 8 g M E T e A F v 4 N 1 6 t l 6 t D + t z V r p i l T 1 H Y A 7 W 1 y 9 + V Z z + < / l a t e x i t > ( 1 )",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" Z n S Z L 6 Y A I 0 t 2 E L 8 7 t 1 4 2 8 X 7 n a f k = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H X K q 5 R X g J H F o k I q S 5 U g J G C r Y G E s E q G V m i h y X K e 1 a s e R 7 S B V U X + B h V 9 h Y Q D E y s j G 3 + C 0 G W j L k S w f n X O v 7 r 0 n S h l V 2 n F + r M r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x 8 8 K J F J T D w s m J D d C C n C a E I 8 T T U j 3 V Q S x C N G O t H o p v A 7 j 0 Q q K p J 7 P U 5 J w N E g o T H F S B s p t B t + J F h f j b n 5 o K / o g C M 4 r 0 V E o 9 A 9 D e 2 6 0 3 S m g M v E L U k d l G i H 9 r f f F z j j J N G Y I a V 6 r p P q I E d S U 8 z I p O Z n i q Q I j 9 C A 9 A x N E C c q y K c X T e C J U f o w F t K 8 R M O p + r c j R 1 w V C 5 p K j v R Q L X q F + J / X y 3 R 8 G e Q 0 S T N N E j w b F G c M a g G L e G C f S o I 1 G x u C s K R m V 4 i H S C K s T Y g 1 E 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "K 7 e P I y 8 c 6 a V 0 3 3 7 r z e u i 7 T q I I j c A w a w A U X o A V u Q R t 4 A I M n 8 A L e w L v 1 b L 1 a H 9 b n r L R i l T 2 H Y A 7 W 1 y 9 J S Z z b < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Z n S Z L 6 Y A I 0 t 2 E L 8 7 t 1 4 2 8 X 7 n a f k = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H X K q 5 R X g J H F o k I q S 5 U g J G C r Y G E s E q G V m i h y X K e 1 a s e R 7 S B V U X + B h V 9 h Y Q D E y s j G 3 + C 0 G W j L k S w f n X O v 7 r 0 n S h l V 2 n F + r M r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x 8 8 K J F J T D w s m J D d C C n C a E I 8 T T U j 3 V Q S x C N G O t H o p v A 7 j 0 Q q K p J 7 P U 5 J w N E g o T H F S B s p t B t + J F h f j b n 5 o K / o g C M 4 r 0 V E o 9 A 9 D e 2 6 0 3 S m g M v E L U k d l G i H 9 r f f F z j j J N G Y I a V 6 r p P q I E d S U 8 z I p O Z n i q Q I j 9 C A 9 A x N E C c q y K c X T e C J U f o w F t K 8 R M O p + r c j R 1 w V C 5 p K j v R Q L X q F + J / X y 3 R 8 G e Q 0 S T N N E j w b F G c M a g G L e G C f S o I 1 G x u C s K R m V 4 i H S C K s T Y g 1 E 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "K 7 e P I y 8 c 6 a V 0 3 3 7 r z e u i 7 T q I I j c A w a w A U X o A V u Q R t 4 A I M n 8 A L e w L v 1 b L 1 a H 9 b n r L R i l T 2 H Y A 7 W 1 y 9 J S Z z b < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Z n S Z L 6 Y A I 0 t 2 E L 8 7 t 1 4 2 8 X 7 n a f k = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H X K q 5 R X g J H F o k I q S 5 U g J G C r Y G E s E q G V m i h y X K e 1 a s e R 7 S B V U X + B h V 9 h Y Q D E y s j G 3 + C 0 G W j L k S w f n X O v 7 r 0 n S h l V 2 n F + r M r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x 8 8 K J F J T D w s m J D d C C n C a E I 8 T T U j 3 V Q S x C N G O t H o p v A 7 j 0 Q q K p J 7 P U 5 J w N E g o T H F S B s p t B t + J F h f j b n 5 o K / o g C M 4 r 0 V E o 9 A 9 D e 2 6 0 3 S m g M v E L U k d l G i H 9 r f f F z j j J N G Y I a V 6 r p P q I E d S U 8 z I p O Z n i q Q I j 9 C A 9 A x N E C c q y K c X T e C J U f o w F t K 8 R M O p + r c j R 1 w V C 5 p K j v R Q L X q F + J / X y 3 R 8 G e Q 0 S T N N E j w b F G c M a g G L e G C f S o I 1 G x u C s K R m V 4 i H S C K s T Y g 1 E 4",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "K 7 e P I y 8 c 6 a V 0 3 3 7 r z e u i 7 T q I I j c A w a w A U X o A V u Q R t 4 A I M n 8 A L e w L v 1 b L 1 a H 9 b n r L R i l T 2 H Y A 7 W 1 y 9 J S Z z b < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Z n S Z L 6 Y A I 0 t 2 E L 8 7 t 1 4 2 8 X 7 n a f k = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "> A A A C E H i c b V C 7 T s M w F H X K q 5 R X g J H F o k I q S 5 U g J G C r Y G E s E q G V m i h y X K e 1 a s e R 7 S B V U X + B h V 9 h Y Q D E y s j G 3 + C 0 G W j L k S w f n X O v 7 r 0 n S h l V 2 n F + r M r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x 8 8 K J F J T D w s m J D d C C n C a E I 8 T T U j 3 V Q S x C N G O t H o p v A 7 j 0 Q q K p J 7 P U 5 J w N E g o T H F S B s p t B t + J F h f j b n 5 o K / o g C M 4 r 0 V E o 9 A 9 D e 2 6 0 3 S m g M v E L U k d l G i H 9 r f f F z j j J N G Y I a V 6 r p P q I E d S U 8 z I p O Z n i q Q I j 9 C A 9 A x N E C c q y K c X T e C J U f o w F t K 8 R M O p + r c j R 1 w V C 5 p K j v R Q L X q F + J / X y 3 R 8 G e Q 0 S T N N E j w b F G c M a g G L e G C f S o I 1 G x u C s K R m V 4 i H S C K s T Y g 1 E 4 K 7 e P I y 8 c 6 a V 0 3 3 7 r z e u i 7 T q I I j c A w a w A U X o A V u Q R t 4 A I M n 8 A L e w L v 1 b L 1 a H 9 b n r L R i l T 2 H Y A 7 W 1 y 9 J S Z z b < / l a t e x i t > \u00b5( 1 )",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" c T H b l G",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T i W 5 x j L 8 a M G w / i Z V U v P E k = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E I r N V H k O G 5 r 1 Y 4 j 2 0 G q o n 4 B C 7 / C w g C I l Z 2 N v 8 F p M 0 D L k S w f n X O v 7 r 0 n S h l V 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x / c K 5 F J T D w s m J C 9 C C n C a E I 8 T T U j v V Q S x C N G u t H 4 u v C 7 D 0 Q q K p I 7 P U l J w N E w o Q O K k T Z S a D f 8 S L B Y T b j 5 o M 8 z 2 P w j R E S j 0 D 0 J 7 b r T c m a A y 8 Q t S R 2 U 6 I T 2 l x 8 L n H G S a M y Q U n 3 X S X W Q I 6 k p Z m R a 8 z N F U o T H a E j 6 h i a I E x X k s 3 O m s G G U G A 6 E N C / R c K b + 7 s g R V 8 W C p p I j P V K L X i H + 5 / U z P b g I c p q k m S Y J n g 8 a Z A x q A Y t s Y E w l w Z p N D E F Y U r M r x C M k E d Y m w Z o J w V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "0 8 e Z l 4 p 6 3 L l n t 7 V m 9 f l W l U w R E 4 B k 3 g g n P Q B j e g A z y A w S N 4 B q / g z X q y X q x 3 6 2 N e W r H K n k P w B 9 b n D + I f m 4 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" c T H b l G",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T i W 5 x j L 8 a M G w / i Z V U v P E k = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E I r N V H k O G 5 r 1 Y 4 j 2 0 G q o n 4 B C 7 / C w g C I l Z 2 N v 8 F p M 0 D L k S w f n X O v 7 r 0 n S h l V 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x / c K 5 F J T D w s m J C 9 C C n C a E I 8 T T U j v V Q S x C N G u t H 4 u v C 7 D 0 Q q K p I 7 P U l J w N E w o Q O K k T Z S a D f 8 S L B Y T b j 5 o M 8 z 2 P w j R E S j 0 D 0 J 7 b r T c m a A y 8 Q t S R 2 U 6 I T 2 l x 8 L n H G S a M y Q U n 3 X S X W Q I 6 k p Z m R a 8 z N F U o T H a E j 6 h i a I E x X k s 3 O m s G G U G A 6 E N C / R c K b + 7 s g R V 8 W C p p I j P V K L X i H + 5 / U z P b g I c p q k m S Y J n g 8 a Z A x q A Y t s Y E w l w Z p N D E F Y U r M r x C M k E d Y m w Z o J w V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "0 8 e Z l 4 p 6 3 L l n t 7 V m 9 f l W l U w R E 4 B k 3 g g n P Q B j e g A z y A w S N 4 B q / g z X q y X q x 3 6 2 N e W r H K n k P w B 9 b n D + I f m 4 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" c T H b l G",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T i W 5 x j L 8 a M G w / i Z V U v P E k = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E I r N V H k O G 5 r 1 Y 4 j 2 0 G q o n 4 B C 7 / C w g C I l Z 2 N v 8 F p M 0 D L k S w f n X O v 7 r 0 n S h l V 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x / c K 5 F J T D w s m J C 9 C C n C a E I 8 T T U j v V Q S x C N G u t H 4 u v C 7 D 0 Q q K p I 7 P U l J w N E w o Q O K k T Z S a D f 8 S L B Y T b j 5 o M 8 z 2 P w j R E S j 0 D 0 J 7 b r T c m a A y 8 Q t S R 2 U 6 I T 2 l x 8 L n H G S a M y Q U n 3 X S X W Q I 6 k p Z m R a 8 z N F U o T H a E j 6 h i a I E x X k s 3 O m s G G U G A 6 E N C / R c K b + 7 s g R V 8 W C p p I j P V K L X i H + 5 / U z P b g I c p q k m S Y J n g 8 a Z A x q A Y t s Y E w l w Z p N D E F Y U r M r x C M k E d Y m w Z o J w V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "0 8 e Z l 4 p 6 3 L l n t 7 V m 9 f l W l U w R E 4 B k 3 g g n P Q B j e g A z y A w S N 4 B q / g z X q y X q x 3 6 2 N e W r H K n k P w B 9 b n D + I f m 4 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" c T H b l G",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T i W 5 x j L 8 a M G w / i Z V U v P E k = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E I r N V H k O G 5 r 1 Y 4 j 2 0 G q o n 4 B C 7 / C w g C I l Z 2 N v 8 F p M 0 D L k S w f n X O v 7 r 0 n S h l V 2 n G + r c r K 6 t r 6 R n W z t r W 9 s 7 t n 7 x / c K 5 F J T D w s m J C 9 C C n C a E I 8 T T U j v V Q S x C N G u t H 4 u v C 7 D 0 Q q K p I 7 P U l J w N E w o Q O K k T Z S a D f 8 S L B Y T b j 5 o M 8 z 2 P w j R E S j 0 D 0 J 7 b r T c m a A y 8 Q t S R 2 U 6 I T 2 l x 8 L n H G S a M y Q U n 3 X S X W Q I 6 k p Z m R a 8 z N F U o T H a E j 6 h i a I E x X k s 3 O m s G G U G A 6 E N C / R c K b + 7 s g R V 8 W C p p I j P V K L X i H + 5 / U z P b g I c p q k m S Y J n g 8 a Z A x q A Y t s Y E w l w Z p N D E F Y U r M r x C M k E d Y m w Z o J w V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "0 8 e Z l 4 p 6 3 L l n t 7 V m 9 f l W l U w R E 4 B k 3 g g n P Q B j e g A z y A w S N 4 B q / g z X q y X q x 3 6 2 N e W r H K n k P w B 9 b n D + I f m 4 4 = < / l a t e x i t >",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "\u00b5( T )",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" s 1 f r",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x K R A s j I e o o F r 6 O d 9 4 v p 6 o c A = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y p I Z W a q L I c Z z W q h N H t o N U R f 0 C F n 6 F h Q E Q K z s b f 4 P T Z q A t R 7 J 8 d M 6 9 u v e e I G V U K s v 6 M S p r 6 x u b W 9 X t 2 s 7 u 3 v 6 B e X j 0 I H k m M H E w Z 1 z 0 A y Q J o w l x F F W M 9 F N B U B w w 0 g v G t 4 X f e y R C U p 5 0 1 S Q l X o y G C Y 0 o R k p L v t l w A 8 5 C O Y n 1 B 9 0 4 g 8 0 F I S A K + d 0 z 3 6 x b L W s G u E r s k t R B i Y 5 v f r s h x 1 l M E o U Z k n J g W 6 n y c i Q U x Y x M a 2 4 m S Y r w G A 3 J Q N M E x U R 6 + e y c K W x o J Y Q R F / o l C s 7 U v x 0 5 i m W x o K 6 M k R r J Z a 8 Q / / M G m Y q u v J w m a a Z I g u e D o o x B x W G R D Q y p I F i x i S Y I C 6 p 3 h X i E B M J K J 1 j T I d j L J 6 8 S 5 7 x 1 3 b L v L + r t m z K N K j g B p 6 A J b H A J 2 u A O d I A D M H g C L + A N v B v P x q v x Y X z O S y t G 2 X M M F m B 8 / Q I X O p u x < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" s 1 f r x K R A s j I e o o F r 6 O d 9 4 v p 6 o c A = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y p I Z W a q L I c Z z W q h N H t o N U R f 0 C F n 6 F h Q E Q K z s b f 4 P T Z q A t R 7 J 8 d M 6 9 u v e e I G V U K s v 6 M S p r 6 x u b W 9 X t 2 s 7 u 3 v 6 B e X j 0 I H k m M H E w Z 1 z 0 A y Q J o w l x F F W M 9 F N B U B w w 0 g v G t 4 X f e y R C U p 5 0 1 S Q l X o y G C Y 0 o R k p L v t l w A 8 5 C O Y n 1 B 9 0 4 g 8 0 F I S A K + d 0 z 3 6 x b L W s G u E r s k t R B i Y 5 v f r s h x 1 l M E o U Z k n J g W 6 n y c i Q U x Y x M a 2 4 m S Y r w G A 3 J Q N M E x U R 6 + e y c K W x o J Y Q R F / o l C s 7 U v x 0 5 i m W x o K 6 M k R r J Z a 8 Q / / M G m Y q u v J w m a a Z I g u e D o o x B x W G R D Q y p I F i x i S Y I C 6 p 3 h X i E B M J K J 1 j T I d j L J 6 8 S 5 7 x 1 3 b L v L + r t m z K N K j g B p 6 A J b H A J 2 u A O d I A D M H g C L + A N v B v P x q v x Y X z O S y t G 2 X M M F m B 8 / Q I X O p u x < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" s 1 f r x K R A s j I e o o F r 6 O d 9 4 v p 6 o c A = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y p I Z W a q L I c Z z W q h N H t o N U R f 0 C F n 6 F h Q E Q K z s b f 4 P T Z q A t R 7 J 8 d M 6 9 u v e e I G V U K s v 6 M S p r 6 x u b W 9 X t 2 s 7 u 3 v 6 B e X j 0 I H k m M H E w Z 1 z 0 A y Q J o w l x F F W M 9 F N B U B w w 0 g v G t 4 X f e y R C U p 5 0 1 S Q l X o y G C Y 0 o R k p L v t l w A 8 5 C O Y n 1 B 9 0 4 g 8 0 F I S A K + d 0 z 3 6 x b L W s G u E r s k t R B i Y 5 v f r s h x 1 l M E o U Z k n J g W 6 n y c i Q U x Y x M a 2 4 m S Y r w G A 3 J Q N M E x U R 6 + e y c K W x o J Y Q R F / o l C s 7 U v x 0 5 i m W x o K 6 M k R r J Z a 8 Q / / M G m Y q u v J w m a a Z I g u e D o o x B x W G R D Q y p I F i x i S Y I C 6 p 3 h X i E B M J K J 1 j T I d j L J 6 8 S 5 7 x 1 3 b L v L + r t m z K N K j g B p 6 A J b H A J 2 u A O d I A D M H g C L + A N v B v P x q v x Y X z O S y t G 2 X M M F m B 8 / Q I X O p u x < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" s 1 f r x K R A s j I e o o F r 6 O d 9 4 v p 6 o c A = \" > A A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y p I Z W a q L I c Z z W q h N H t o N U R f 0 C F n 6 F h Q E Q K z s b f 4 P T Z q A t R 7 J 8 d M 6 9 u v e e I G V U K s v 6 M S p r 6 x u b W 9 X t 2 s 7 u 3 v 6 B e X j 0 I H k m M H E w Z 1 z 0 A y Q J o w l x F F W M 9 F N B U B w w 0 g v G t 4 X f e y R C U p 5 0 1 S Q l X o y G C Y 0 o R k p L v t l w A 8 5 C O Y n 1 B 9 0 4 g 8 0 F I S A K + d 0 z 3 6 x b L W s G u E r s k t R B i Y 5 v f r s h x 1 l M E o U Z k n J g W 6 n y c i Q U x Y x M a 2 4 m S Y r w G A 3 J Q N M E x U R 6 + e y c K W x o J Y Q R F / o l C s 7 U v x 0 5 i m W x o K 6 M k R r J Z a 8 Q / / M G m Y q u v J w m a a Z I g u e D o o x B x W G R D Q y p I F i x i S Y I C 6 p 3 h X i E B M J K J 1 j T I d j L J 6 8 S 5 7 x 1 3 b L v L + r t m z K N K j g B p 6 A J b H A J 2 u A O d I A D M H g C L + A N v B v P x q v x Y X z O S y t G 2 X M M F m B 8 / Q I X O p u x < / l a t e x i t > \u00b5 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" G N t L L E C x T g 9 u 9 M u z K F X q o 3 m q / M 0 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X j x O s G 6 w l p K m 6 R a W J i V J l V H 3 U b x 4 U P H q N / H m t z H d e t D N B y G P 9 3 4 / 8 v K i j F G l H e f b q q 2 s r q 1 v 1 D c b W 9 s 7 u 3 t 2 c / 9 e i V x i 4 m H B h O x H S B F G O f E 0 1 Y z 0 M 0 l Q G j H S i 8 b X p d 9 7 I F J R w e / 0 J C N B i o a c J h Q j b a T Q b v q R Y L G a p O a C f p q H b m i 3 n L Y z A 1 w m b k V a o E I 3 t L / 8 W O A 8 J V x j h p Q a u E 6 m g w J J T T E j 0 4 a f K 5 I h P E Z D M j C U o 5 S o o J h F n 8 J j o 8 Q w E d I c r u F M / b 1 R o F S V 6 c x k i v R I L X q l + J 8 3 y H V y E R S U Z 7 k m H M 8 f S n I G t Y B l D z C m k m D N J o Y g L K n J C v E I S Y S 1 a a t h S n A X v 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x M v N P 2 Z d u 9 P W t 1 r q o 2 6 u A Q H I E T 4 I J z 0 A E 3 o A s 8 g M E j e A a v 4 M 1 6 s l 6 s d + t j P l q z q p 0 D 8 A f W 5 w 9 y X 5 O q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" G N t L L E C x T g 9 u 9 M u z K F X q o 3 m q / M 0 = \" > A A A B + X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X j x O s G 6 w l p K m 6 R a W J i V J l V H 3 U b x 4 U P H q N / H m t z H d e t D N B y G P 9 3 4 / 8 v K i j F G l H e f b q q 2 s r q 1 v 1 D c b W 9 s 7 u 3 t 2 c / 9 e i V x i 4 m H B h O x H S B F G O f E 0 1 Y z 0 M 0 l Q G j H S i 8 b X p d 9 7 I F J R w e / 0 J C N B i o a c J h Q j b a T Q b v q R Y L G a p O a C f p q H b m i 3 n L Y z A 1 w m b k V a o E I 3 t L / 8 W O A 8 J V x j h p Q a u E 6 m g w J J T T E j 0 4 a f K 5 I h P E Z D M j C U o 5 S o o J h F n 8 J j o 8 Q w E d I c r u F M / b 1 R o F S V 6 c x k i v R I L X q l + J 8 3 y H V y E R S U Z 7 k m H M 8 f S n I G t Y B l D z C m k m D N J o Y g L K n J C v E I S Y S 1 a a t h S n A X v 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x M v N P 2 Z d u 9 P W t 1 r q o 2 6 u A Q H I E T 4 I J z 0 A E 3 o A s 8 g M E j e A a v 4 M 1 6 s l 6 s d + t j P l q z q p 0 D 8 A f W 5 w 9 y X 5 O q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" G N t L L E C x T g 9 u 9 M u z K F X q o 3 m q / M 0 = \" > A A A B + X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X j x O s G 6 w l p K m 6 R a W J i V J l V H 3 U b x 4 U P H q N / H m t z H d e t D N B y G P 9 3 4 / 8 v K i j F G l H e f b q q 2 s r q 1 v 1 D c b W 9 s 7 u 3 t 2 c / 9 e i V x i 4 m H B h O x H S B F G O f E 0 1 Y z 0 M 0 l Q G j H S i 8 b X p d 9 7 I F J R w e / 0 J C N B i o a c J h Q j b a T Q b v q R Y L G a p O a C f p q H b m i 3 n L Y z A 1 w m b k V a o E I 3 t L / 8 W O A 8 J V x j h p Q a u E 6 m g w J J T T E j 0 4 a f K 5 I h P E Z D M j C U o 5 S o o J h F n 8 J j o 8 Q w E d I c r u F M / b 1 R o F S V 6 c x k i v R I L X q l + J 8 3 y H V y E R S U Z 7 k m H M 8 f S n I G t Y B l D z C m k m D N J o Y g L K n J C v E I S Y S 1 a a t h S n A X v 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x M v N P 2 Z d u 9 P W t 1 r q o 2 6 u A Q H I E T 4 I J z 0 A E 3 o A s 8 g M E j e A a v 4 M 1 6 s l 6 s d + t j P l q z q p 0 D 8 A f W 5 w 9 y X 5 O q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" G N t L L E C x T g 9 u 9 M u z K F X q o 3 m q / M 0 = \" > A A A B + X i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X j x O s G 6 w l p K m 6 R a W J i V J l V H 3 U b x 4 U P H q N / H m t z H d e t D N B y G P 9 3 4 / 8 v K i j F G l H e f b q q 2 s r q 1 v 1 D c b W 9 s 7 u 3 t 2 c / 9 e i V x i 4 m H B h O x H S B F G O f E 0 1 Y z 0 M 0 l Q G j H S i 8 b X p d 9 7 I F J R w e / 0 J C N B i o a c J h Q j b a T Q b v q R Y L G a p O a C f p q H b m i 3 n L Y z A 1 w m b k V a o E I 3 t L / 8 W O A 8 J V x j h p Q a u E 6 m g w J J T T E j 0 4 a f K 5 I h P E Z D M j C U o 5 S o o J h F n 8 J j o 8 Q w E d I c r u F M / b 1 R o F S V 6 c x k i v R I L X q l + J 8 3 y H V y E R S U Z 7 k m H M 8 f S n I G t Y B l D z C m k m D N J o Y g L K n J C v E I S Y S 1 a a t h S n A X v 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x M v N P 2 Z d u 9 P W t 1 r q o 2 6 u A Q H I E T 4 I J z 0 A E 3 o A s 8 g M E j e A a v 4 M 1 6 s l 6 s d + t j P l q z q p 0 D 8 A f W 5 w 9 y X 5 O q < / l a t e x i t > \u00b5 T < l a t e x i t s h a 1 _ b a s e 6 4 = \" J T P a x 7 n c V 7 K s N 3 l i w R K x T z W l H M A = \" > A A A B + X i c b V B P S 8 M w H E 3 9 O + e / T o 9 e g k P w",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "N F o R 1 N v Q i 8 c J q x u s p a R p u o W l S U l S Z d R 9 F C 8 e V L z 6 T b z 5 b U y 3 H n T z Q c j j v d + P v L w o Y 1 R p x / m 2 V l b X 1 j c 2 a 1 v 1 7 Z 3 d v X 2 7 c X C v R C 4 x 8 b B g Q v Y j p A i j n H i a a k b 6 m S Q o j R j p R e O b 0 u 8 9 E K m o 4 F 0 9 y U i Q o i G n C c V I G y m 0 G 3 4 k W K w m q b m g n + Z h N 7 S b T s u Z A S 4 T t y J N U K E T 2 l 9 + L H C e E q 4 x Q 0 o N X C f T Q Y G k p p i R a d 3 P F c k Q H q M h G R j K U U p U U M y i T + G J U W K Y C G k O 1 3 C m / t 4 o U K r K d G Y y R X q k F r 1 S / M 8 b 5 D q 5 D A r K s 1 w T j u c P J T m D W s C y B x h T S b B m E 0 M Q l t R k h X i E J M L a t F U 3 J b i L X 1 4 m 3 l n r q u X e n T f b 1 1 U b N X A E j s E p c M E F a I N b 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A E e w O A R P I N X 8 G Y 9 W S / W u / U x H 1 2 x q p 1 D 8 A f W 5 w + n S J P N < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" J T P a x 7 n c V 7 K s N 3 l i w R K x T z W l H M A = \" > A A A B + X i c b V B P S 8 M w H E 3 9 O + e / T o 9 e g k P w",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "N F o R 1 N v Q i 8 c J q x u s p a R p u o W l S U l S Z d R 9 F C 8 e V L z 6 T b z 5 b U y 3 H n T z Q c j j v d + P v L w o Y 1 R p x / m 2 V l b X 1 j c 2 a 1 v 1 7 Z 3 d v X 2 7 c X C v R C 4 x 8 b B g Q v Y j p A i j n H i a a k b 6 m S Q o j R j p R e O b 0 u 8 9 E K m o 4 F 0 9 y U i Q o i G n C c V I G y m 0 G 3 4 k W K w m q b m g n + Z h N 7 S b T s u Z A S 4 T t y J N U K E T 2 l 9 + L H C e E q 4 x Q 0 o N X C f T Q Y G k p p i R a d 3 P F c k Q H q M h G R j K U U p U U M y i T + G J U W K Y C G k O 1 3 C m / t 4 o U K r K d G Y y R X q k F r 1 S / M 8 b 5 D q 5 D A r K s 1 w T j u c P J T m D W s C y B x h T S b B m E 0 M Q l t R k h X i E J M L a t F U 3 J b i L X 1 4 m 3 l n r q u X e n T f b 1 1 U b N X A E j s E p c M E F a I N b 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A E e w O A R P I N X 8 G Y 9 W S / W u / U x H 1 2 x q p 1 D 8 A f W 5 w + n S J P N < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" J T P a x 7 n c V 7 K s N 3 l i w R K x T z W l H M A = \" > A A A B + X i c b V B P S 8 M w H E 3 9 O + e / T o 9 e g k P w",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "N F o R 1 N v Q i 8 c J q x u s p a R p u o W l S U l S Z d R 9 F C 8 e V L z 6 T b z 5 b U y 3 H n T z Q c j j v d + P v L w o Y 1 R p x / m 2 V l b X 1 j c 2 a 1 v 1 7 Z 3 d v X 2 7 c X C v R C 4 x 8 b B g Q v Y j p A i j n H i a a k b 6 m S Q o j R j p R e O b 0 u 8 9 E K m o 4 F 0 9 y U i Q o i G n C c V I G y m 0 G 3 4 k W K w m q b m g n + Z h N 7 S b T s u Z A S 4 T t y J N U K E T 2 l 9 + L H C e E q 4 x Q 0 o N X C f T Q Y G k p p i R a d 3 P F c k Q H q M h G R j K U U p U U M y i T + G J U W K Y C G k O 1 3 C m / t 4 o U K r K d G Y y R X q k F r 1 S / M 8 b 5 D q 5 D A r K s 1 w T j u c P J T m D W s C y B x h T S b B m E 0 M Q l t R k h X i E J M L a t F U 3 J b i L X 1 4 m 3 l n r q u X e n T f b 1 1 U b N X A E j s E p c M E F a I N b 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A E e w O A R P I N X 8 G Y 9 W S / W u / U x H 1 2 x q p 1 D 8 A f W 5 w + n S J P N < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" J T P a x 7 n c V 7 K s N 3 l i w R K x T z W l H M A = \" > A A A B + X i c b V B P S 8 M w H E 3 9 O + e / T o 9 e g k P w",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "N F o R 1 N v Q i 8 c J q x u s p a R p u o W l S U l S Z d R 9 F C 8 e V L z 6 T b z 5 b U y 3 H n T z Q c j j v d + P v L w o Y 1 R p x / m 2 V l b X 1 j c 2 a 1 v 1 7 Z 3 d v X 2 7 c X C v R C 4 x 8 b B g Q v Y j p A i j n H i a a k b 6 m S Q o j R j p R e O b 0 u 8 9 E K m o 4 F 0 9 y U i Q o i G n C c V I G y m 0 G 3 4 k W K w m q b m g n + Z h N 7 S b T s u Z A S 4 T t y J N U K E T 2 l 9 + L H C e E q 4 x Q 0 o N X C f T Q Y G k p p i R a d 3 P F c k Q H q M h G R j K U U p U U M y i T + G J U W K Y C G k O 1 3 C m / t 4 o U K r K d G Y y R X q k F r 1 S / M 8 b 5 D q 5 D A r K s 1 w T j u c P J T m D W s C y B x h T S b B m E 0 M Q l t R k h X i E J M L a t F U 3 J b i L X 1 4 m 3 l n r q u X e n T f b 1 1 U b N X A E j s E p c M E F a I N b 0 A E e w O A R P I N X 8 G Y 9 W S / W u / U x H 1 2 x q p 1 D 8 A f W 5 w + n S J P N < / l a t e x i t > 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" I + 3 9 O 6 t a a H d y 5 Z c 4 5 b 8 z i G b r Y P w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V Q b 0 N v X i c Y N 1 g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e v R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i p B u N r k u / + 0 C k o i K 9 0 + O M B B w N U p p Q j L S R Q v v A j w S L 1 Z i b C / q K D j g K 3 d B u O i 1 n C r h I 3 I o 0 Q Y V O a H / 5 s c A 5 J 6 n G D C n V d 5 1 M B w W S m m J G J g 0 / V y R D e I Q G p G 9 o i j h R Q T F N P 4 H H R o l h I q Q 5 q Y Z T 9 f d G g b g q A 5 p J j v R Q z X u l + J / X z 3 V y E R Q 0 z X J N U j x 7 K M k Z 1 A K W V c C Y S o I 1 G x u C s K Q m K 8 R D J B H W p r C G K c G d / / I i 8 U 5 b l y 3 3 9 q z Z v q r a q I N D c A R O g A v O Q R v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "c g A 7 w A A a P 4 B m 8 g j f r y X q x 3 q 2 P 2 W j N q n b 2 w R 9 Y n z / B V p T 3 < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" I + 3 9 O 6 t a a H d y 5 Z c 4 5 b 8 z i G b r Y P w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V Q b 0 N v X i c Y N 1 g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e v R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i p B u N r k u / + 0 C k o i K 9 0 + O M B B w N U p p Q j L S R Q v v A j w S L 1 Z i b C / q K D j g K 3 d B u O i 1 n C r h I 3 I o 0 Q Y V O a H / 5 s c A 5 J 6 n G D C n V d 5 1 M B w W S m m J G J g 0 / V y R D e I Q G p G 9 o i j h R Q T F N P 4 H H R o l h I q Q 5 q Y Z T 9 f d G g b g q A 5 p J j v R Q z X u l + J / X z 3 V y E R Q 0 z X J N U j x 7 K M k Z 1 A K W V c C Y S o I 1 G x u C s K Q m K 8 R D J B H W p r C G K c G d / / I i 8 U 5 b l y 3 3 9 q z Z v q r a q I N D c A R O g A v O Q R v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "c g A 7 w A A a P 4 B m 8 g j f r y X q x 3 q 2 P 2 W j N q n b 2 w R 9 Y n z / B V p T 3 < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" I + 3 9 O 6 t a a H d y 5 Z c 4 5 b 8 z i G b r Y P w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V Q b 0 N v X i c Y N 1 g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e v R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i p B u N r k u / + 0 C k o i K 9 0 + O M B B w N U p p Q j L S R Q v v A j w S L 1 Z i b C / q K D j g K 3 d B u O i 1 n C r h I 3 I o 0 Q Y V O a H / 5 s c A 5 J 6 n G D C n V d 5 1 M B w W S m m J G J g 0 / V y R D e I Q G p G 9 o i j h R Q T F N P 4 H H R o l h I q Q 5 q Y Z T 9 f d G g b g q A 5 p J j v R Q z X u l + J / X z 3 V y E R Q 0 z X J N U j x 7 K M k Z 1 A K W V c C Y S o I 1 G x u C s K Q m K 8 R D J B H W p r C G K c G d / / I i 8 U 5 b l y 3 3 9 q z Z v q r a q I N D c A R O g A v O Q R v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "c g A 7 w A A a P 4 B m 8 g j f r y X q x 3 q 2 P 2 W j N q n b 2 w R 9 Y n z / B V p T 3 < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" I + 3 9 O 6 t a a H d y 5 Z c 4 5 b 8 z i G b r Y P w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V Q b 0 N v X i c Y N 1 g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e v R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i p B u N r k u / + 0 C k o i K 9 0 + O M B B w N U p p Q j L S R Q v v A j w S L 1 Z i b C / q K D j g K 3 d B u O i 1 n C r h I 3 I o 0 Q Y V O a H / 5 s c A 5 J 6 n G D C n V d 5 1 M B w W S m m J G J g 0 / V y R D e I Q G p G 9 o i j h R Q T F N P 4 H H R o l h I q Q 5 q Y Z T 9 f d G g b g q A 5 p J j v R Q z X u l + J / X z 3 V y E R Q 0 z X J N U j x 7 K M k Z 1 A K W V c C Y S o I 1 G x u C s K Q m K 8 R D J B H W p r C G K c G d / / I i 8 U 5 b l y 3 3 9 q z Z v q r a q I N D c A R O g A v O Q R v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "c g A 7 w A A a P 4 B m 8 g j f r y X q x 3 q 2 P 2 W j N q n b 2 w R 9 Y n z / B V p T 3 < / l a t e x i t > T < l a t e x i t s h a 1 _ b a s e 6 4 = \" a m p h J B I Y k 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "J i 2 I m o Q q X Z C E u C t O w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M V Q b 0 N v X i c s O p g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e n R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i 5 D 4 a X Z f + / Q O R i o q 0 q 8 c Z C T g a p D S h G G k j h f a B H w k W q z E 3 F / Q V H X A U d k O 7 6 b S c K e A i c S v S B B U 6 o f 3 l x w L n n K Q a M 6 R U 3 3 U y H R R I a o o Z m T T 8 X J E M 4 R E a k L 6 h K e J E B c U 0 / Q Q e G y W G i Z D m p B p O 1 d 8 b B e K q D G g m O d J D N e + V 4 n 9 e P 9 f J R V D Q N M s 1 S f H s o S R n U A t Y V g F j K g n W b G w I w p K a r B A P k U R Y m 8 I a p g R 3 / s u L x D t t X b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b c 2 7 N m + 6 p q o w 4 O w R E 4 A S 4 4 B 2 1 w A z r A A x g 8 g m f w C t 6 s J + v F e r c + Z q M 1 q 9 r Z B 3 9 g f f 4 A 9 j + V G g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" a m p h J B I Y k 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "J i 2 I m o Q q X Z C E u C t O w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M V Q b 0 N v X i c s O p g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e n R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i 5 D 4 a X Z f + / Q O R i o q 0 q 8 c Z C T g a p D S h G G k j h f a B H w k W q z E 3 F / Q V H X A U d k O 7 6 b S c K e A i c S v S B B U 6 o f 3 l x w L n n K Q a M 6 R U 3 3 U y H R R I a o o Z m T T 8 X J E M 4 R E a k L 6 h K e J E B c U 0 / Q Q e G y W G i Z D m p B p O 1 d 8 b B e K q D G g m O d J D N e + V 4 n 9 e P 9 f J R V D Q N M s 1 S f H s o S R n U A t Y V g F j K g n W b G w I w p K a r B A P k U R Y m 8 I a p g R 3 / s u L x D t t X b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b c 2 7 N m + 6 p q o w 4 O w R E 4 A S 4 4 B 2 1 w A z r A A x g 8 g m f w C t 6 s J + v F e r c + Z q M 1 q 9 r Z B 3 9 g f f 4 A 9 j + V G g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" a m p h J B I Y k 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "J i 2 I m o Q q X Z C E u C t O w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M V Q b 0 N v X i c s O p g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e n R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i 5 D 4 a X Z f + / Q O R i o q 0 q 8 c Z C T g a p D S h G G k j h f a B H w k W q z E 3 F / Q V H X A U d k O 7 6 b S c K e A i c S v S B B U 6 o f 3 l x w L n n K Q a M 6 R U 3 3 U y H R R I a o o Z m T T 8 X J E M 4 R E a k L 6 h K e J E B c U 0 / Q Q e G y W G i Z D m p B p O 1 d 8 b B e K q D G g m O d J D N e + V 4 n 9 e P 9 f J R V D Q N M s 1 S f H s o S R n U A t Y V g F j K g n W b G w I w p K a r B A P k U R Y m 8 I a p g R 3 / s u L x D t t X b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b c 2 7 N m + 6 p q o w 4 O w R E 4 A S 4 4 B 2 1 w A z r A A x g 8 g m f w C t 6 s J + v F e r c + Z q M 1 q 9 r Z B 3 9 g f f 4 A 9 j + V G g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" a m p h J B I Y k 0",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "J i 2 I m o Q q X Z C E u C t O w = \" > A A A B / H i c b V D N S 8 M w H E 3 n 1 5 x f 9 e P m J T g E T 6 M V Q b 0 N v X i c s O p g L S V N 0 y 0 s a U q S C r M M / x U v H l S 8 + o d 4 8 7 8 x 3 X r Q z Q c h j / d + P / L y o o x R p R 3 n 2 6 o t L a + s r t X X G x u b W 9 s 7 9 u 7 e n R K 5 x M T D g g n Z i 5 A i j K b E 0 1 Q z 0 s s k Q T x i 5 D 4 a X Z f + / Q O R i o q 0 q 8 c Z C T g a p D S h G G k j h f a B H w k W q z E 3 F / Q V H X A U d k O 7 6 b S c K e A i c S v S B B U 6 o f 3 l x w L n n K Q a M 6 R U 3 3 U y H R R I a o o Z m T T 8 X J E M 4 R E a k L 6 h K e J E B c U 0 / Q Q e G y W G i Z D m p B p O 1 d 8 b B e K q D G g m O d J D N e + V 4 n 9 e P 9 f J R V D Q N M s 1 S f H s o S R n U A t Y V g F j K g n W b G w I w p K a r B A P k U R Y m 8 I a p g R 3 / s u L x D t t X b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b c 2 7 N m + 6 p q o w 4 O w R E 4 A S 4 4 B 2 1 w A z r A A x g 8 g m f w C t 6 s J + v F e r c + Z q M 1 q 9 r Z B 3 9 g f f 4 A 9 j + V G g = = < / l a t e x i t > z 0 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9 b n D / I h k t E = < / l a t e x i t > z 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3 R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9 b n D / O k k t I = < / l a t e x i t > z K 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y V a 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "/ M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y V a 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "/ M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y V a 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "/ M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y V a 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6 / M H z i 2 U a g = = < / l a t e x i t > z K",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" n U",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > v1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" r 8 c 1 m F V n 1 l i Q p C / A U a 1 B p e a v m O I = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 Y e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J 0 s E s + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Q c q Z 0 r b 9 b V X W 1 j c 2 t 6 r b t Z",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "3 d v f 1 6 4 + D w U S W Z J N Q l C U 9 k L 8 C K c i a o q 5 n m t J d K i u O A 0 2 4 w v i 3 8 7 o R K x R L x o K c p 9 W I 8 F C x i B G s j + Y 3 6 I E h 4 q K a x u d D E d / x G 0 2 7 Z c 6 B V 4 p S k C S U 6 f u N r E C Y k i 6 n Q h G O l + o 6 d a i / H U j P C 6 a w 2 y B R N M R n j I e 0 b K n B M l Z f P g 8 / Q q V F C F C X S H K H R X P 2 9 k e N Y F d n M Z I z 1 S C 1 7 h f i f 1 8 9 0 d O X l T K S Z p o I s H o o y j n S C i h Z Q y C Q l m k 8 N w U Q y k x W R E Z a Y a N N V z Z T g L H 9 5 l b j n r e u W c 3 / R b N + U b V T h G E 7 g D B y 4 h D b c Q Q d c I J D B M 7 z C m / V k v V j v 1 s d i t G K V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "O 0 f w B 9 b n D + 2 Q k s 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" r 8 c 1 m F V n 1 l i Q p C / A U a 1 B p e a v m O I = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 Y e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J 0 s E s + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Q c q Z 0 r b 9 b V X W 1 j c 2 t 6 r b t Z",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "3 d v f 1 6 4 + D w U S W Z J N Q l C U 9 k L 8 C K c i a o q 5 n m t J d K i u O A 0 2 4 w v i 3 8 7 o R K x R L x o K c p 9 W I 8 F C x i B G s j + Y 3 6 I E h 4 q K a x u d D E d / x G 0 2 7 Z c 6 B V 4 p S k C S U 6 f u N r E C Y k i 6 n Q h G O l + o 6 d a i / H U j P C 6 a w 2 y B R N M R n j I e 0 b K n B M l Z f P g 8 / Q q V F C F C X S H K H R X P 2 9 k e N Y F d n M Z I z 1 S C 1 7 h f i f 1 8 9 0 d O X l T K S Z p o I s H o o y j n S C i h Z Q y C Q l m k 8 N w U Q y k x W R E Z a Y a N N V z Z T g L H 9 5 l b j n r e u W c 3 / R b N + U b V T h G E 7 g D B y 4 h D b c Q Q d c I J D B M 7 z C m / V k v V j v 1 s d i t G K V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "O 0 f w B 9 b n D + 2 Q k s 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" r 8 c 1 m F V n 1 l i Q p C / A U a 1 B p e a v m O I = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 Y e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J 0 s E s + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Q c q Z 0 r b 9 b V X W 1 j c 2 t 6 r b t Z",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "3 d v f 1 6 4 + D w U S W Z J N Q l C U 9 k L 8 C K c i a o q 5 n m t J d K i u O A 0 2 4 w v i 3 8 7 o R K x R L x o K c p 9 W I 8 F C x i B G s j + Y 3 6 I E h 4 q K a x u d D E d / x G 0 2 7 Z c 6 B V 4 p S k C S U 6 f u N r E C Y k i 6 n Q h G O l + o 6 d a i / H U j P C 6 a w 2 y B R N M R n j I e 0 b K n B M l Z f P g 8 / Q q V F C F C X S H K H R X P 2 9 k e N Y F d n M Z I z 1 S C 1 7 h f i f 1 8 9 0 d O X l T K S Z p o I s H o o y j n S C i h Z Q y C Q l m k 8 N w U Q y k x W R E Z a Y a N N V z Z T g L H 9 5 l b j n r e u W c 3 / R b N + U b V T h G E 7 g D B y 4 h D b c Q Q d c I J D B M 7 z C m / V k v V j v 1 s d i t G K V",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "O 0 f w B 9 b n D + 2 Q k s 4 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" r 8 c 1 m F V n 1 l i Q p C / A U a 1 B p e a v m O I = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 Y e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J 0 s E s + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Q c q Z 0 r b 9 b V X W 1 j c 2 t 6 r b t Z 3 d v f 1 6 4 + D w U S W Z J N Q l C U 9 k L 8 C K c i a o q 5 n m t J d K i u O A 0 2 4 w v i 3 8 7 o R K x R L x o K c p 9 W I 8 F C x i B G s j + Y 3 6 I E h 4 q K a x u d D E d / x G 0 2 7 Z c 6 B V 4 p S k C S U 6 f u N r E C Y k i 6 n Q h G O l + o 6 d a i / H U j P C 6 a w 2 y B R N M R n j I e 0 b K n B M l Z f P g 8 / Q q V F C F C X S H K H R X P 2 9 k e N Y F d n M Z I z 1 S C 1 7 h f i f 1 8 9 0 d O X l T K S Z p o I s H o o y j n S C i h Z Q y C Q l m k 8 N w U Q y k x W R E Z a Y a N N V z Z T g L H 9 5 l b j n r e u W c 3 / R b N + U b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "V T h G E 7 g D B y 4 h D b c Q Q d c I J D B M 7 z C m / V k v V j v 1 s d i t G K V O 0 f w B 9 b n D + 2 Q k s 4 = < / l a t e x i t > v2",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" + I E 4 L T X q V a 0 E F z 1 s z 1 j 6 M 9 M n r U 0 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 2 C O p t 6 M X j B O s G W y l p m m 5 h a V q S d D D H P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 2 v r G 5 t b 2 5 W d 6 u 7 e / k H N P j x 6 V G k u C f V I y l P Z D b G i n A n q a a Y 5 7 W a S 4 i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T k t B O O b g u / M 6 Z S s V Q 8 6 E l G / Q Q P B I s Z w d p I g V 3 r h y m P 1 C Q x F x o H z c C u O w 1 n D r R K 3 J L U o U Q 7 s L / 6 U U r y h A p N O F a q 5 z q Z 9 q d Y a k Y 4 n V X 7 u a I Z J i M 8 o D 1 D B U 6 o 8 q f z 4 D N 0 Z p Q I x a k 0 R 2 g 0 V 3 9 v T H G i i m x m M s F 6 q J a 9 Q v z P 6 + U 6 v v K n T G S 5 p o I s H o p z j n S K i h Z Q x C Q l m k 8 M w U Q y k x W R I Z a Y a N N V 1 Z T g L n 9 5 l X j N x n X D v b + o t 2 7 K N i p w A q d w D i 5 c Q g v u o",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A 0 e E M j h G V 7 h z X q y X q x 3 6 2 M x u m a V O 8 f w B 9 b n D + 8 T k s 8 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" + I E 4 L T X q V a 0 E F z 1 s z 1 j 6 M 9 M n r U 0 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 2 C O p t 6 M X j B O s G W y l p m m 5 h a V q S d D D H P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 2 v r G 5 t b 2 5 W d 6 u 7 e / k H N P j x 6 V G k u C f V I y l P Z D b G i n A n q a a Y 5 7 W a S 4 i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T k t B O O b g u / M 6 Z S s V Q 8 6 E l G / Q Q P B I s Z w d p I g V 3 r h y m P 1 C Q x F x o H z c C u O w 1 n D r R K 3 J L U o U Q 7 s L / 6 U U r y h A p N O F a q 5 z q Z 9 q d Y a k Y 4 n V X 7 u a I Z J i M 8 o D 1 D B U 6 o 8 q f z 4 D N 0 Z p Q I x a k 0 R 2 g 0 V 3 9 v T H G i i m x m M s F 6 q J a 9 Q v z P 6 + U 6 v v K n T G S 5 p o I s H o p z j n S K i h Z Q x C Q l m k 8 M w U Q y k x W R I Z a Y a N N V 1 Z T g L n 9 5 l X j N x n X D v b + o t 2 7 K N i p w A q d w D i 5 c Q g v u o",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A 0 e E M j h G V 7 h z X q y X q x 3 6 2 M x u m a V O 8 f w B 9 b n D + 8 T k s 8 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" + I E 4 L T X q V a 0 E F z 1 s z 1 j 6 M 9 M n r U 0 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 2 C O p t 6 M X j B O s G W y l p m m 5 h a V q S d D D H P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 2 v r G 5 t b 2 5 W d 6 u 7 e / k H N P j x 6 V G k u C f V I y l P Z D b G i n A n q a a Y 5 7 W a S 4 i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T k t B O O b g u / M 6 Z S s V Q 8 6 E l G / Q Q P B I s Z w d p I g V 3 r h y m P 1 C Q x F x o H z c C u O w 1 n D r R K 3 J L U o U Q 7 s L / 6 U U r y h A p N O F a q 5 z q Z 9 q d Y a k Y 4 n V X 7 u a I Z J i M 8 o D 1 D B U 6 o 8 q f z 4 D N 0 Z p Q I x a k 0 R 2 g 0 V 3 9 v T H G i i m x m M s F 6 q J a 9 Q v z P 6 + U 6 v v K n T G S 5 p o I s H o p z j n S K i h Z Q x C Q l m k 8 M w U Q y k x W R I Z a Y a N N V 1 Z T g L n 9 5 l X j N x n X D v b + o t 2 7 K N i p w A q d w D i 5 c Q g v u o",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A 0 e E M j h G V 7 h z X q y X q x 3 6 2 M x u m a V O 8 f w B 9 b n D + 8 T k s 8 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" + I E 4 L T X q V a 0 E F z 1 s z 1 j 6 M 9 M n r U 0 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 2 C O p t 6 M X j B O s G W y l p m m 5 h a V q S d D D H P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 2 v r G 5 t b 2 5 W d 6 u 7 e / k H N P j x 6 V G k u C f V I y l P Z D b G i n A n q a a Y 5 7 W a S 4 i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "T k t B O O b g u / M 6 Z S s V Q 8 6 E l G / Q Q P B I s Z w d p I g V 3 r h y m P 1 C Q x F x o H z c C u O w 1 n D r R K 3 J L U o U Q 7 s L / 6 U U r y h A p N O F a q 5 z q Z 9 q d Y a k Y 4 n V X 7 u a I Z J i M 8 o D 1 D B U 6 o 8 q f z 4 D N 0 Z p Q I x a k 0 R 2 g 0 V 3 9 v T H G i i m x m M s F 6 q J a 9 Q v z P 6 + U 6 v v K n T G S 5 p o I s H o p z j n S K i h Z Q x C Q l m k 8 M w U Q y k x W R I Z a Y a N N V 1 Z T g L n 9 5 l X j N x n X D v b + o t 2 7 K N i p w A q d w D i 5 c Q g v u o",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A 0 e E M j h G V 7 h z X q y X q x 3 6 2 M x u m a V O 8 f w B 9 b n D + 8 T k s 8 = < / l a t e x i t > vK < l a t e x i t s h a 1 _ b a s e 6 4 = \" g I q l O O z r B k G u v h g J + M V K j Y h G 9 v 4 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 W B P U 2 9 C J 4 m W D d Y C s l T d M t L G 1 K k g 7 m 2 C f x 4 k H F q 1 / F m 9 / G d O t B N x + E P N 7 7 / c j L C z P O l H a c b 2 t l d W 1 9 Y 7 O y V d 3 e 2 d 2 r 2 f s H j 0 r k k l C P C C 5 k J 8 S K c p Z S T z P N a S e T F C c h p + 1 w e F P 4 7 R G V i o n 0 Q Y 8 z 6 i e 4 n 7 K Y E a y N F N i 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "X i h 4 p M a J u d A o u A v s u t N w Z k D L x C 1 J H U q 0 A v u r F w m S J z T V h G O l u q 6 T a X + C p W a E 0 2 m 1 l y u a Y T L E f d o 1 N M U J V f 5 k F n y K T o w S o V h I c 1 K N Z u r v j Q l O V J H N T C Z Y D 9 S i V 4 j / e d 1 c x 5 f + h K V Z r m l K 5 g / F O U d a o K I F F D F J i e Z j Q z C R z G R F Z I A l J t p 0 V T U l u I t f X i b e W e O q 4 d 6 f 1 5 v X Z R s V O I J j O A U X L q A J t 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A C D w j k 8 A y v 8 G Y 9 W S / W u / U x H 1 2 x y p 1 D + A P r 8 w c U 7 Z L o < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" g I q l O O z r B k G u v h g J + M V K j Y h G 9 v 4 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 W B P U 2 9 C J 4 m W D d Y C s l T d M t L G 1 K k g 7 m 2 C f x 4 k H F q 1 / F m 9 / G d O t B N x + E P N 7 7 / c j L C z P O l H a c b 2 t l d W 1 9 Y 7 O y V d 3 e 2 d 2 r 2 f s H j 0 r k k l C P C C 5 k J 8 S K c p Z S T z P N a S e T F C c h p + 1 w e F P 4 7 R G V i o n 0 Q Y 8 z 6 i e 4 n 7 K Y E a y N F N i 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "X i h 4 p M a J u d A o u A v s u t N w Z k D L x C 1 J H U q 0 A v u r F w m S J z T V h G O l u q 6 T a X + C p W a E 0 2 m 1 l y u a Y T L E f d o 1 N M U J V f 5 k F n y K T o w S o V h I c 1 K N Z u r v j Q l O V J H N T C Z Y D 9 S i V 4 j / e d 1 c x 5 f + h K V Z r m l K 5 g / F O U d a o K I F F D F J i e Z j Q z C R z G R F Z I A l J t p 0 V T U l u I t f X i b e W e O q 4 d 6 f 1 5 v X Z R s V O I J j O A U X L q A J t 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A C D w j k 8 A y v 8 G Y 9 W S / W u / U x H 1 2 x y p 1 D + A P r 8 w c U 7 Z L o < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" g I q l O O z r B k G u v h g J + M V K j Y h G 9 v 4 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 W B P U 2 9 C J 4 m W D d Y C s l T d M t L G 1 K k g 7 m 2 C f x 4 k H F q 1 / F m 9 / G d O t B N x + E P N 7 7 / c j L C z P O l H a c b 2 t l d W 1 9 Y 7 O y V d 3 e 2 d 2 r 2 f s H j 0 r k k l C P C C 5 k J 8 S K c p Z S T z P N a S e T F C c h p + 1 w e F P 4 7 R G V i o n 0 Q Y 8 z 6 i e 4 n 7 K Y E a y N F N i 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "X i h 4 p M a J u d A o u A v s u t N w Z k D L x C 1 J H U q 0 A v u r F w m S J z T V h G O l u q 6 T a X + C p W a E 0 2 m 1 l y u a Y T L E f d o 1 N M U J V f 5 k F n y K T o w S o V h I c 1 K N Z u r v j Q l O V J H N T C Z Y D 9 S i V 4 j / e d 1 c x 5 f + h K V Z r m l K 5 g / F O U d a o K I F F D F J i e Z j Q z C R z G R F Z I A l J t p 0 V T U l u I t f X i b e W e O q 4 d 6 f 1 5 v X Z R s V O I J j O A U X L q A J t 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A C D w j k 8 A y v 8 G Y 9 W S / W u / U x H 1 2 x y p 1 D + A P r 8 w c U 7 Z L o < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" g I q l O O z r B k G u v h g J + M V K j Y h G 9 v 4 = \" > A A A B 9 3 i c b V B P S 8 M w H P 3 V v 3 P + W d W j l + A Q P I 1 W B P U 2 9 C J 4 m W D d Y C s l T d M t L G 1 K k g 7 m 2 C f x 4 k H F q 1 / F m 9 / G d O t B N x + E P N 7 7 / c j L C z P O l H a c b 2 t l d W 1 9 Y 7 O y V d 3 e 2 d 2 r 2 f s H j 0 r k k l C P C C 5 k J 8 S K c p Z S T z P N a S e T F C c h p + 1 w e F P 4 7 R G V i o n 0 Q Y 8 z 6 i e 4 n 7 K Y E a y N F N i 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "X i h 4 p M a J u d A o u A v s u t N w Z k D L x C 1 J H U q 0 A v u r F w m S J z T V h G O l u q 6 T a X + C p W a E 0 2 m 1 l y u a Y T L E f d o 1 N M U J V f 5 k F n y K T o w S o V h I c 1 K N Z u r v j Q l O V J H N T C Z Y D 9 S i V 4 j / e d 1 c x 5 f + h K V Z r m l K 5 g / F O U d a o K I F F D F J i e Z j Q z C R z G R F Z I A l J t p 0 V T U l u I t f X i b e W e O q 4 d 6 f 1 5 v X Z R s V O I J j O A U X L q A J t 9 A C D w j k 8 A y v 8 G Y 9 W S / W u / U x H 1 2 x y p 1 D + A P r 8 w c U 7 Z L o < / l a t e x i t > p(d|\u2713) < l",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Encoder Decoder",
                "sec_num": null
            },
            {
                "text": "A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E o r N V H l O E 5 r 1 X n I v k G q Q r + A h V 9 h Y Q D E y s 7 G 3 + C 0 G U r L k S w f n 3 O v f O / x E s E V W N a P U V p Z X V v f K G 9 W t r Z 3 d v f M / Y N 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "F a e S s j a N R S y 7 H l F M 8 I i 1 g Y N g 3 U Q y E n q C d b z R d e 5 3 H p h U P I 7 u Y J w w N y S D i A e c E t B S 3 6 w l d c e L h a / G o b 6 w j x / x / N u B I Q N y 0 j e r V s O a A i 8 T u y B V V K D V N 7 8 d P 6 Z p y C K g g i j V s 6 0 E 3 I x I 4 F S w S c V J F U s I H Z E B 6 2 k a k Z A p N 5 u u M 8 E 1 r f g 4 i K U + E e C p O t + R k V D l 8 + n K k M B Q L X q 5 + J / X S y G 4 c D M e J S m w i M 4 + C l K B I c Z 5 N t j n k l E Q Y 0 0 I l V z P i u m Q S E J B J 1 j R I d i L K y + T 9 m n j s m H f n l W b V 0 U a Z X S E j l E d 2 e g c N d E N a q E 2 o u g J v a A 3 9 G 4 8 G 6 / G h / E 5 K y 0 Z R c 8 h + g P j 6 x c G i Z u q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E o r N V H l O E 5 r 1 X n I v k G q Q r + A h V 9 h Y Q D E y s 7 G 3 + C 0 G U r L k S w f n 3 O v f O / x E s E V W N a P U V p Z X V v f K G 9 W t r Z 3 d v f M / Y N 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "F a e S s j a N R S y 7 H l F M 8 I i 1 g Y N g 3 U Q y E n q C d b z R d e 5 3 H p h U P I 7 u Y J w w N y S D i A e c E t B S 3 6 w l d c e L h a / G o b 6 w j x / x / N u B I Q N y 0 j e r V s O a A i 8 T u y B V V K D V N 7 8 d P 6 Z p y C K g g i j V s 6 0 E 3 I x I 4 F S w S c V J F U s I H Z E B 6 2 k a k Z A p N 5 u u M 8 E 1 r f g 4 i K U + E e C p O t + R k V D l 8 + n K k M B Q L X q 5 + J / X S y G 4 c D M e J S m w i M 4 + C l K B I c Z 5 N t j n k l E Q Y 0 0 I l V z P i u m Q S E J B J 1 j R I d i L K y + T 9 m n j s m H f n l W b V 0 U a Z X S E j l E d 2 e g c N d E N a q E 2 o u g J v a A 3 9 G 4 8 G 6 / G h / E 5 K y 0 Z R c 8 h + g P j 6 x c G i Z u q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E o r N V H l O E 5 r 1 X n I v k G q Q r + A h V 9 h Y Q D E y s 7 G 3 + C 0 G U r L k S w f n 3 O v f O / x E s E V W N a P U V p Z X V v f K G 9 W t r Z 3 d v f M / Y N 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "F a e S s j a N R S y 7 H l F M 8 I i 1 g Y N g 3 U Q y E n q C d b z R d e 5 3 H p h U P I 7 u Y J w w N y S D i A e c E t B S 3 6 w l d c e L h a / G o b 6 w j x / x / N u B I Q N y 0 j e r V s O a A i 8 T u y B V V K D V N 7 8 d P 6 Z p y C K g g i j V s 6 0 E 3 I x I 4 F S w S c V J F U s I H Z E B 6 2 k a k Z A p N 5 u u M 8 E 1 r f g 4 i K U + E e C p O t + R k V D l 8 + n K k M B Q L X q 5 + J / X S y G 4 c D M e J S m w i M 4 + C l K B I c Z 5 N t j n k l E Q Y 0 0 I l V z P i u m Q S E J B J 1 j R I d i L K y + T 9 m n j s m H f n l W b V 0 U a Z X S E j l E d 2 e g c N d E N a q E 2 o u g J v a A 3 9 G 4 8 G 6 / G h / E 5 K y 0 Z R c 8 h + g P j 6 x c G i Z u q < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "A A C D X i c b V C 7 T s M w F H X K q 5 R X g J H F o q p U l i p B S M B W w c J Y J E o r N V H l O E 5 r 1 X n I v k G q Q r + A h V 9 h Y Q D E y s 7 G 3 + C 0 G U r L k S w f n 3 O v f O / x E s E V W N a P U V p Z X V v f K G 9 W t r Z 3 d v f M / Y N 7",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "F a e S s j a N R S y 7 H l F M 8 I i 1 g Y N g 3 U Q y E n q C d b z R d e 5 3 H p h U P I 7 u Y J w w N y S D i A e c E t B S 3 6 w l d c e L h a / G o b 6 w j x / x / N u B I Q N y 0 j e r V s O a A i 8 T u y B V V K D V N 7 8 d P 6 Z p y C K g g i j V s 6 0 E 3 I x I 4 F S w S c V J F U s I H Z E B 6 2 k a k Z A p N 5 u u M 8 E 1 r f g 4 i K U + E e C p O t + R k V D l 8 + n K k M B Q L X q 5 + J / X S y G 4 c D M e J S m w i M 4 + C l K B I c Z 5 N t j n k l E Q Y 0 0 I l V z P i u m Q S E J B J 1 j R I d i L K y + T 9 m n j s m H f n l W b V 0 U a Z X S E j l E d 2 e g c N d E N a q E 2 o u g J v a A 3 9 G 4 8 G 6 / G h / E 5 K y 0 Z R c 8 h + g P j 6 x c G i Z u q < / l a t e x i t > q(\u2713|d)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" 0 e Z X F w + f r f d u 4 u z S M e X l Z 8 R U G y o = \" > A A A C D n i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V C 7 T s M w F H V 4 l v I K M L J Y V K C y V A l C A r Y K F s Y i E V q p j S r H c V q r z g P 7 B q k K / Q M W f o W F A R A r M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x t / g 9 N m K C 1 H s n x 0 z r 2 6 9 x 4 v E V y B Z f 0 Y C 4 t L y y u r p b X y + s b m 1 r a 5 s 3 u n 4 l R S 5 t B Y x L L l E c U E j 5 g D H A R r J Z K R 0 B O s 6 Q 2 u c r / 5 w K T i c X Q L w 4 S 5 I e l F P O C U g J a 6 5 t E 9 r n a 8 W P h q G O o P d 6 D P g O B H P C 3 6 x 1 2 z Y t W s M f A 8 s Q t S Q Q U a X f O 7 4 8 c 0 D V k E V B C l 2 r a V g J s R C Z w K N i p 3 U s U S Q g e k x 9 q a R i R k y s 3 G 9 4 z w o V Z 8 H M R S v w j w W J 3 u y E i o 8 t V 0 Z U i g r 2 a 9 X P z P a 6 c Q n L s Z j 5 I U W E Q n g 4 J U Y I h x H g 7 2 u W Q U x F A T Q i X X u 2 L a J 5 J Q 0 B G W d Q j 2 7 M n z x D m p X d T s m 9 N K / b J I o 4 T 2 0 Q G q I h u d o T q 6 R g 3 k I I q e 0 A t 6 Q + / G s / F q f B i f k 9 I F o + j Z Q 3 9 g f P 0 C Z o y b 1 Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 0 e Z X F w + f r f d u 4 u z S M e X l Z 8 R U G y o = \" > A A A C D n i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V C 7 T s M w F H V 4 l v I K M L J Y V K C y V A l C A r Y K F s Y i E V q p j S r H c V q r z g P 7 B q k K / Q M W f o W F A R A r M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x t / g 9 N m K C 1 H s n x 0 z r 2 6 9 x 4 v E V y B Z f 0 Y C 4 t L y y u r p b X y + s b m 1 r a 5 s 3 u n 4 l R S 5 t B Y x L L l E c U E j 5 g D H A R r J Z K R 0 B O s 6 Q 2 u c r / 5 w K T i c X Q L w 4 S 5 I e l F P O C U g J a 6 5 t E 9 r n a 8 W P h q G O o P d 6 D P g O B H P C 3 6 x 1 2 z Y t W s M f A 8 s Q t S Q Q U a X f O 7 4 8 c 0 D V k E V B C l 2 r a V g J s R C Z w K N i p 3 U s U S Q g e k x 9 q a R i R k y s 3 G 9 4 z w o V Z 8 H M R S v w j w W J 3 u y E i o 8 t V 0 Z U i g r 2 a 9 X P z P a 6 c Q n L s Z j 5 I U W E Q n g 4 J U Y I h x H g 7 2 u W Q U x F A T Q i X X u 2 L a J 5 J Q 0 B G W d Q j 2 7 M n z x D m p X d T s m 9 N K / b J I o 4 T 2 0 Q G q I h u d o T q 6 R g 3 k I I q e 0 A t 6 Q + / G s / F q f B i f k 9 I F o + j Z Q 3 9 g f P 0 C Z o y b 1 Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 0 e Z X F w + f r f d u 4 u z S M e X l Z 8 R U G y o = \" > A A A C D n i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V C 7 T s M w F H V 4 l v I K M L J Y V K C y V A l C A r Y K F s Y i E V q p j S r H c V q r z g P 7 B q k K / Q M W f o W F A R A r M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x t / g 9 N m K C 1 H s n x 0 z r 2 6 9 x 4 v E V y B Z f 0 Y C 4 t L y y u r p b X y + s b m 1 r a 5 s 3 u n 4 l R S 5 t B Y x L L l E c U E j 5 g D H A R r J Z K R 0 B O s 6 Q 2 u c r / 5 w K T i c X Q L w 4 S 5 I e l F P O C U g J a 6 5 t E 9 r n a 8 W P h q G O o P d 6 D P g O B H P C 3 6 x 1 2 z Y t W s M f A 8 s Q t S Q Q U a X f O 7 4 8 c 0 D V k E V B C l 2 r a V g J s R C Z w K N i p 3 U s U S Q g e k x 9 q a R i R k y s 3 G 9 4 z w o V Z 8 H M R S v w j w W J 3 u y E i o 8 t V 0 Z U i g r 2 a 9 X P z P a 6 c Q n L s Z j 5 I U W E Q n g 4 J U Y I h x H g 7 2 u W Q U x F A T Q i X X u 2 L a J 5 J Q 0 B G W d Q j 2 7 M n z x D m p X d T s m 9 N K / b J I o 4 T 2 0 Q G q I h u d o T q 6 R g 3 k I I q e 0 A t 6 Q + / G s / F q f B i f k 9 I F o + j Z Q 3 9 g f P 0 C Z o y b 1 Q = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 0 e Z X F w + f r f d u 4 u z S M e X l Z 8 R U G y o = \" > A A A C D n i c b",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V C 7 T s M w F H V 4 l v I K M L J Y V K C y V A l C A r Y K F s Y i E V q p j S r H c V q r z g P 7 B q k K / Q M W f o W F A R A r M",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x t / g 9 N m K C 1 H s n x 0 z r 2 6 9 x 4 v E V y B Z f 0 Y C 4 t L y y u r p b X y + s b m 1 r a 5 s 3 u n 4 l R S 5 t B Y x L L l E c U E j 5 g D H A R r J Z K R 0 B O s 6 Q 2 u c r / 5 w K T i c X Q L w 4 S 5 I e l F P O C U g J a 6 5 t E 9 r n a 8 W P h q G O o P d 6 D P g O B H P C 3 6 x 1 2 z Y t W s M f A 8 s Q t S Q Q U a X f O 7 4 8 c 0 D V k E V B C l 2 r a V g J s R C Z w K N i p 3 U s U S Q g e k x 9 q a R i R k y s 3 G 9 4 z w o V Z 8 H M R S v w j w W J 3 u y E i o 8 t V 0 Z U i g r 2 a 9 X P z P a 6 c Q n L s Z j 5 I U W E Q n g 4 J U Y I h x H g 7 2 u W Q U x F A T Q i X X u 2 L a J 5 J Q 0 B G W d Q j 2 7 M n z x D m p X d T s m 9 N K / b J I o 4 T 2 0 Q G q I h u d o T q 6 R g 3 k I I q e 0 A t 6 Q + / G s / F q f B i f k 9 I F o + j Z Q 3 9 g f P 0 C Z o y b 1 Q = = < / l a t e x i t > Householder Flow \u21e1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" 8 N / I Q b O F a u n n 9 A 9 U N g P v r j N g V J 8 = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w O 1 j L S N N 3 C 0 q Y k q T D L P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 7 W V 1 b X 1 j f p m Y 2 t 7 Z 7 d p 7 + 3 f K 5 F L Q j 0 i u J C 9 E C v K W U o 9 z T S n v U x S n I S c P o T j 6 9 J / e K R S M Z H e 6 U l G g w Q P U x Y z g r W R B n b T D w W P 1 C Q x F / I z N r B b T t u Z A S 0 T t y I t q N A d 2 F 9 + J E i e 0 F Q T j p X q u 0 6 m g w J L z Q i n 0 4 a f",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "K 5 p h M s Z D 2 j c 0 x Q l V Q T E L P k X H R o l Q L K Q 5 q U Y z 9 f d G g R N V Z j O T C d Y j t e i V 4 n 9 e P 9 f x R V C w N M s 1 T c n 8 o T j n S A t U t o A i J i n R f G I I J p K Z r I i M s M R E m 6 4 a p g R 3 8 c v L",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x D t t X 7 b d 2 7 N W 5 6 p q o w 6 H c A Q n 4 M I 5 d O A G u u A B g R y e 4 R X e r C f r x X q 3 P u a j N a v a O Y A / s D 5 / A D S J k v 0 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 8 N / I Q b O F a u n n 9 A 9 U N g P v r j N g V J 8 = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w O 1 j L S N N 3 C 0 q Y k q T D L P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 7 W V 1 b X 1 j f p m Y 2 t 7 Z 7 d p 7 + 3 f K 5 F L Q j 0 i u J C 9 E C v K W U o 9 z T S n v U x S n I S c P o T j 6 9 J / e K R S M Z H e 6 U l G g w Q P U x Y z g r W R B n b T D w W P 1 C Q x F / I z N r B b T t u Z A S 0 T t y I t q N A d 2 F 9 + J E i e 0 F Q T j p X q u 0 6 m g w J L z Q i n 0 4 a f",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "K 5 p h M s Z D 2 j c 0 x Q l V Q T E L P k X H R o l Q L K Q 5 q U Y z 9 f d G g R N V Z j O T C d Y j t e i V 4 n 9 e P 9 f x R V C w N M s 1 T c n 8 o T j n S A t U t o A i J i n R f G I I J p K Z r I i M s M R E m 6 4 a p g R 3 8 c v L",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x D t t X 7 b d 2 7 N W 5 6 p q o w 6 H c A Q n 4 M I 5 d O A G u u A B g R y e 4 R X e r C f r x X q 3 P u a j N a v a O Y A / s D 5 / A D S J k v 0 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 8 N / I Q b O F a u n n 9 A 9 U N g P v r j N g V J 8 = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w O 1 j L S N N 3 C 0 q Y k q T D L P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 7 W V 1 b X 1 j f p m Y 2 t 7 Z 7 d p 7 + 3 f K 5 F L Q j 0 i u J C 9 E C v K W U o 9 z T S n v U x S n I S c P o T j 6 9 J / e K R S M Z H e 6 U l G g w Q P U x Y z g r W R B n b T D w W P 1 C Q x F / I z N r B b T t u Z A S 0 T t y I t q N A d 2 F 9 + J E i e 0 F Q T j p X q u 0 6 m g w J L z Q i n 0 4 a f",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "K 5 p h M s Z D 2 j c 0 x Q l V Q T E L P k X H R o l Q L K Q 5 q U Y z 9 f d G g R N V Z j O T C d Y j t e i V 4 n 9 e P 9 f x R V C w N M s 1 T c n 8 o T j n S A t U t o A i J i n R f G I I J p K Z r I i M s M R E m 6 4 a p g R 3 8 c v L",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x D t t X 7 b d 2 7 N W 5 6 p q o w 6 H c A Q n 4 M I 5 d O A G u u A B g R y e 4 R X e r C f r x X q 3 P u a j N a v a O Y A / s D 5 / A D S J k v 0 = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" 8 N / I Q b O F a u n n 9 A 9 U N g P v r j N g V J 8 = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w O 1 j L S N N 3 C 0 q Y k q T D L P o k X D y p e / S r e / D a m W w + 6 + S D k 8 d 7 v R 1 5 e m H G m t O N 8 W 7 W V 1 b X 1 j f p m Y 2 t 7 Z 7 d p 7 + 3 f K 5 F L Q j 0 i u J C 9 E C v K W U o 9 z T S n v U x S n I S c P o T j 6 9 J / e K R S M Z H e 6 U l G g w Q P U x Y z g r W R B n b T D w W P 1 C Q x F / I z N r B b T t u Z A S 0 T t y I t q N A d 2 F 9 + J E i e 0 F Q T j p X q u 0 6 m g w J L z Q i n 0 4 a f ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "K 5 p h M s Z D 2 j c 0 x Q l V Q T E L P k X H R o l Q L K Q 5 q U Y z 9 f d G g R N V Z j O T C d Y j t e i V 4 n 9 e P 9 f x R V C w N M s 1 T c n 8 o T j n S A t U t o A i J i n R f G I I J p K Z r I i M s M R E m 6 4 a p g R 3 8 c v L x D t t X",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V R z G o e S t 8 L R z d R v P X J t R K L u c Z z y I K Y D J S L B K F r J f + r l 7 q R X r b l 1 d w a y T L y C 1 K B A s 1 f 9 6 v Y T l s V c I Z P U m I 7 n p h j k V K N g k k 8 q 3 c z w l L I R H f C O p Y r G 3 A T 5 7 N g J O b F K n 0 S J t q W Q z N T f E z m N j R n H o e 2 M K Q 7 N o j c V / / M 6 G U a X Q S 5 U m i F X b L 4 o y i T B h E w / J 3 2 h O U M 5 t o Q y L e y t h A 2 p p g x t P h U b g r f 4 8 j L x z + p X d e / u v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V R z G o e S t 8 L R z d R v P X J t R K L u c Z z y I K Y D J S L B K F r J f + r l 7 q R X r b l 1 d w a y T L y C 1 K B A s 1 f 9 6 v Y T l s V c I Z P U m I 7 n p h j k V K N g k k 8 q 3 c z w l L I R H f C O p Y r G 3 A T 5 7 N g J O b F K n 0 S J t q W Q z N T f E z m N j R n H o e 2 M K Q 7 N o j c V / / M 6 G U a X Q S 5 U m i F X b L 4 o y i T B h E w / J 3 2 h O U M 5 t o Q y L e y t h A 2 p p g x t P h U b g r f 4 8 j L x z + p X d e / u v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V R z G o e S t 8 L R z d R v P X J t R K L u c Z z y I K Y D J S L B K F r J f + r l 7 q R X r b l 1 d w a y T L y C 1 K B A s 1 f 9 6 v Y T l s V c I Z P U m I 7 n p h j k V K N g k k 8 q 3 c z w l L I R H f C O p Y r G 3 A T 5 7 N g J O b F K n 0 S J t q W Q z N T f E z m N j R n H o e 2 M K Q 7 N o j c V / / M 6 G U a X Q S 5 U m i F X b L 4 o y i T B h E w / J 3 2 h O U M 5 t o Q y L e y t h A 2 p p g x t P h U b g r f 4 8 j L x z + p X d e / u v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V R z G o e S t 8 L R z d R v P X J t R K L u c Z z y I K Y D J S L B K F r J f + r l 7 q R X r b l 1 d w a y T L y C 1 K B A s 1 f 9 6 v Y T l s V c I Z P U m I 7 n p h j k V K N g k k 8 q 3 c z w l L I R H f C O p Y r G 3 A T 5 7 N g J O b F K n 0 S J t q W Q z N T f E z m N j R n H o e 2 M K Q 7 N o j c V / / M 6 G U a X Q S 5 U m i F X b L 4 o y i T B h E w / J 3 2 h O U M 5 t o Q y L e y t h A 2 p p g x t P h U b g r f 4 8 j L x z + p X d e / u v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "i 8 Q r S B U K N L q V r 0 4 v Y V n M F T J J j W l 7 b o p B T j U K J v m 4 3 M k M T y k b 0 j 5 v W 6 p o z E 2 Q T 4 8 d k 2 O r 9 E i U a F s K y V T 9 P Z H T 2 J h R H N r O m O L A z H s T 8 T + v n W F 0 E e R C p R l y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x W a L o k w S T M j k c 9 I T m j O U I 0 s o 0 8 L e S t i A a s r Q 5 l O 2 I X j z L y 8 S / 7 R 2 W f P u z q r 1 q y K N E h z C E Z y A B + d Q h x t o g A 8 M B D z D K 7 w 5 y n l x 3 p 2 P W e u S U 8 w c w B 8 4 n z 9 p F I 6 d < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W x + R J U e i k j N u 4 4 x i q x C 2 D v c D j 8 o = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x Z N U M L b Q h r L Z b t q l m 0 3 Y n Y g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m E p h 0 H W / n a X l l d W 1 9 d J G e X N r e 2 e 3 s r f / Y J J M M + 6 z R C a 6 F V L D p V D c R 4 G S t 1 L N a R x K 3 g y H 1 x O / + c i 1 E Y m 6 x 1 H K g 5 j 2 l Y g E o 2 g l / 6 m b 3 4 6 7 l a p b c 6 c g i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "8 Q r S B U K N L q V r 0 4 v Y V n M F T J J j W l 7 b o p B T j U K J v m 4 3 M k M T y k b 0 j 5 v W 6 p o z E 2 Q T 4 8 d k 2 O r 9 E i U a F s K y V T 9 P Z H T 2 J h R H N r O m O L A z H s T 8 T + v n W F 0 E e R C p R l y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x W a L o k w S T M j k c 9 I T m j O U I 0 s o 0 8 L e S t i A a s r Q 5 l O 2 I X j z L y 8 S / 7 R 2 W f P u z q r 1 q y K N E h z C E Z y A B + d Q h x t o g A 8 M B D z D K 7 w 5 y n l x 3 p 2 P W e u S U 8 w c w B 8 4 n z 9 p F I 6 d < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W x + R J U e i k j N u 4 4 x i q x C 2 D v c D j 8 o = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x Z N U M L b Q h r L Z b t q l m 0 3 Y n Y g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m E p h 0 H W / n a X l l d W 1 9 d J G e X N r e 2 e 3 s r f / Y J J M M + 6 z R C a 6 F V L D p V D c R 4 G S t 1 L N a R x K 3 g y H 1 x O / + c i 1 E Y m 6 x 1 H K g 5 j 2 l Y g E o 2 g l / 6 m b 3 4 6 7 l a p b c 6 c g i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "8 Q r S B U K N L q V r 0 4 v Y V n M F T J J j W l 7 b o p B T j U K J v m 4 3 M k M T y k b 0 j 5 v W 6 p o z E 2 Q T 4 8 d k 2 O r 9 E i U a F s K y V T 9 P Z H T 2 J h R H N r O m O L A z H s T 8 T + v n W F 0 E e R C p R l y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x W a L o k w S T M j k c 9 I T m j O U I 0 s o 0 8 L e S t i A a s r Q 5 l O 2 I X j z L y 8 S / 7 R 2 W f P u z q r 1 q y K N E h z C E Z y A B + d Q h x t o g A 8 M B D z D K 7 w 5 y n l x 3 p 2 P W e u S U 8 w c w B 8 4 n z 9 p F I 6 d < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W x + R J U e i k j N u 4 4 x i q x C 2 D v c D j 8 o = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x Z N U M L b Q h r L Z b t q l m 0 3 Y n Y g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m E p h 0 H W / n a X l l d W 1 9 d J G e X N r e 2 e 3 s r f / Y J J M M + 6 z R C a 6 F V L D p V D c R 4 G S t 1 L N a R x K 3 g y H 1 x O / + c i 1 E Y m 6 x 1 H K g 5 j 2 l Y g E o 2 g l / 6 m b 3 4 6 7 l a p b c 6 c g i",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "8 Q r S B U K N L q V r 0 4 v Y V n M F T J J j W l 7 b o p B T j U K J v m 4 3 M k M T y k b 0 j 5 v W 6 p o z E 2 Q T 4 8 d k 2 O r 9 E i U a F s K y V T 9 P Z H T 2 J h R H N r O m O L A z H s T 8 T + v n W F 0 E e R C p R l y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x W a L o k w S T M j k c 9 I T m j O U I 0 s o 0 8 L e S t i A a s r Q 5 l O 2 I X j z L y 8 S / 7 R 2 W f P u z q r 1 q y K N E h z C E Z y A B + d Q h x t o g A 8 M B D z D K 7 w 5 y n l x 3 p 2 P W e u S U 8 w c w B 8 4 n z 9 p F I 6 d < / l a t e x i t > Decoder GRU GRU GRU y 2 < l a t e x i t s h a 1 _ b a s e 6 4 = \" n r 6 N a O H A t W Z 4 P o Y e / P Y + y J Q c 8 f 4 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K k k R 1 F v R i 8 e K x h b a U D b b T b t 0 s w m 7 E 6 G E / g Q v H l S 8 + o + 8 + W / c t j l o 6 4 O B x 3 s z z M w L U y k M u u 6 3 s 7 K 6 t r 6 x W d o q b + / s 7 u 1 X D g 4 f T Z J p x n 2 W y E S 3 Q 2 q 4 F I r 7 K F D y d q o 5 j U P J W + H o Z u q 3 n r g 2 I l E P O E 5 5 E N O B E p F g F K 1 0 P + 7 V e 5 W q W 3 N n I M v E K 0 g V C j R 7 l a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "i b 5 p N z N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z B T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T t i F 4 i y 8 v E 7 9 e u 6 p 5 d + f V x n W R R g m O 4 Q T O w I M L a M A t N M E H B g N 4 h l d 4 c 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "T z 4 r w 7 H / P W F a e Y O Y I / c D 5 / A H x 9 j X Y = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n r 6 N a O H A t W Z 4 P o Y e / P Y + y J Q c 8 f 4 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K k k R 1 F v R i 8 e K x h b a U D b b T b t 0 s w m 7 E 6 G E / g Q v H l S 8 + o + 8 + W / c t j l o 6 4 O B x 3 s z z M w L U y k M u u 6 3 s 7 K 6 t r 6 x W d o q b + / s 7 u 1 X D g 4 f T Z J p x n 2 W y E S 3 Q 2 q 4 F I r 7 K F D y d q o 5 j U P J W + H o Z u q 3 n r g 2 I l E P O E 5 5 E N O B E p F g F K 1 0 P + 7 V e 5 W q W 3 N n I M v E K 0 g V C j R 7 l a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "i b 5 p N z N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z B T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T t i F 4 i y 8 v E 7 9 e u 6 p 5 d + f V x n W R R g m O 4 Q T O w I M L a M A t N M E H B g N 4 h l d 4 c 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "T z 4 r w 7 H / P W F a e Y O Y I / c D 5 / A H x 9 j X Y = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n r 6 N a O H A t W Z 4 P o Y e / P Y + y J Q c 8 f 4 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K k k R 1 F v R i 8 e K x h b a U D b b T b t 0 s w m 7 E 6 G E / g Q v H l S 8 + o + 8 + W / c t j l o 6 4 O B x 3 s z z M w L U y k M u u 6 3 s 7 K 6 t r 6 x W d o q b + / s 7 u 1 X D g 4 f T Z J p x n 2 W y E S 3 Q 2 q 4 F I r 7 K F D y d q o 5 j U P J W + H o Z u q 3 n r g 2 I l E P O E 5 5 E N O B E p F g F K 1 0 P + 7 V e 5 W q W 3 N n I M v E K 0 g V C j R 7 l a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "i b 5 p N z N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z B T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T t i F 4 i y 8 v E 7 9 e u 6 p 5 d + f V x n W R R g m O 4 Q T O w I M L a M A t N M E H B g N 4 h l d 4 c 6",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "T z 4 r w 7 H / P W F a e Y O Y I / c D 5 / A H x 9 j X Y = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n r 6 N a O H A t W Z 4 P o Y e / P Y + y J Q c 8 f 4 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K k k R 1 F v R i 8 e K x h b a U D b b T b t 0 s w m 7 E 6 G E / g Q v H l S 8 + o + 8 + W / c t j l o 6 4 O B x 3 s z z M w L U y k M u u 6 3 s 7 K 6 t r 6 x W d o q b + / s 7 u 1 X D g 4 f T Z J p x n 2 W y E S 3 Q 2 q 4 F I r 7 K F D y d q o 5 j U P J W + H o Z u q 3 n r g 2 I l E P O E 5 5 E N O B E p F g F K 1 0 P + 7 V e 5 W q W 3 N n I M v E K 0 g V C j R 7 l a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "i b 5 p N z N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z B T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T t i F 4 i y 8 v E 7 9 e u 6 p 5 d + f V x n W R R g m O 4 Q T O w I M L a M A t N M E H B g N",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "N i R T Z / q u 1 L I E V T m q F W b x V y A = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E i s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 E L x 5 U v P q P v P l v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "3 L Y 5 a P X B w O O 9 G W b m h a n g 2 r j u l 1 N a W l 5 Z X S u v V z Y 2 t 7 Z 3 q r t 7 D z r J F E O f J S J R 7 Z B q F F y i b 7 g R 2 E 4 V 0 j g U 2 A p H V 1 O / 9 Y h K 8 0 T e m 3 G K Q U w H k k e c U W O l u 3 H v p l e t u X V 3 B v K X e A W p Q Y F m r / r Z 7 S c s i 1 E a J q j W H c 9 N T Z B T Z T g T O K l 0 M 4 0 p Z S M 6 w I 6 l k s a o g 3 x 2 6 o Q c W a V P o k T Z k o b M 1 J 8 T O Y 2 1 H s e h 7 Y y p G e p F b y r + 5 3 U y E 5 0 H O Z d p Z l C y + a I o E 8 Q k Z P o 3 6 X O F z I i x J Z Q p b m 8 l b E g V Z c a m U 7 E h e I s v / y X + S f 2 i 7 t 2 e 1 h q X R R p l O I B D O A Y P z q A B 1 9 A E H x g M 4 A l e 4 N U R z r P z 5 r z P W 0 t O M b M P v + B 8 f A O l T o 2 R < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R 2 Y A N i R T Z / q u 1 L I E V T m q F W b x V y A = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E i s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 E L x 5 U v P q P v P l v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "3 L Y 5 a P X B w O O 9 G W b m h a n g 2 r j u l 1 N a W l 5 Z X S u v V z Y 2 t 7 Z 3 q r t 7 D z r J F E O f J S J R 7 Z B q F F y i b 7 g R 2 E 4 V 0 j g U 2 A p H V 1 O / 9 Y h K 8 0 T e m 3 G K Q U w H k k e c U W O l u 3 H v p l e t u X V 3 B v K X e A W p Q Y F m r / r Z 7 S c s i 1 E a J q j W H c 9 N T Z B T Z T g T O K l 0 M 4 0 p Z S M 6 w I 6 l k s a o g 3 x 2 6 o Q c W a V P o k T Z k o b M 1 J 8 T O Y 2 1 H s e h 7 Y y p G e p F b y r + 5 3 U y E 5 0 H O Z d p Z l C y + a I o E 8 Q k Z P o 3 6 X O F z I i x J Z Q p b m 8 l b E g V Z c a m U 7 E h e I s v / y X + S f 2 i 7 t 2 e 1 h q X R R p l O I B D O A Y P z q A B 1 9 A E H x g M 4 A l e 4 N U R z r P z 5 r z P W 0 t O M b M P v + B 8 f A O l T o 2 R < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R 2 Y A N i R T Z / q u 1 L I E V T m q F W b x V y A = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E i s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 E L x 5 U v P q P v P l v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "3 L Y 5 a P X B w O O 9 G W b m h a n g 2 r j u l 1 N a W l 5 Z X S u v V z Y 2 t 7 Z 3 q r t 7 D z r J F E O f J S J R 7 Z B q F F y i b 7 g R 2 E 4 V 0 j g U 2 A p H V 1 O / 9 Y h K 8 0 T e m 3 G K Q U w H k k e c U W O l u 3 H v p l e t u X V 3 B v K X e A W p Q Y F m r / r Z 7 S c s i 1 E a J q j W H c 9 N T Z B T Z T g T O K l 0 M 4 0 p Z S M 6 w I 6 l k s a o g 3 x 2 6 o Q c W a V P o k T Z k o b M 1 J 8 T O Y 2 1 H s e h 7 Y y p G e p F b y r + 5 3 U y E 5 0 H O Z d p Z l C y + a I o E 8 Q k Z P o 3 6 X O F z I i x J Z Q p b m 8 l b E g V Z c a m U 7 E h e I s v / y X + S f 2 i 7 t 2 e 1 h q X R R p l O I B D O A Y P z q A B 1 9 A E H x g M 4 A l e 4 N U R z r P z 5 r z P W 0 t O M b M P v + B 8 f A O l T o 2 R < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R 2 Y A N i R T Z / q u 1 L I E V T m q F W b x V y A = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L 1 6 E i s Y W 2 l A 2 2 0 m 7 d L M J u x u h h P 4 E L x 5 U v P q P v P l v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "3 L Y 5 a P X B w O O 9 G W b m h a n g 2 r j u l 1 N a W l 5 Z X S u v V z Y 2 t 7 Z 3 q r t 7 D z r J F E O f J S J R 7 Z B q F F y i b 7 g R 2 E 4 V 0 j g U 2 A p H V 1 O / 9 Y h K 8 0 T e m 3 G K Q U w H k k e c U W O l u 3 H v p l e t u X V 3 B v K X e A W p Q Y F m r / r Z 7 S c s i 1 E a J q j W H c 9 N T Z B T Z T g T O K l 0 M 4 0 p Z S M 6 w I 6 l k s a o g 3 x 2 6 o Q c W a V P o k T Z k o b M 1 J 8 T O Y 2 1 H s e h 7 Y y p G e p F b y r + 5 3 U y E 5 0 H O Z d p Z l C y + a I o E 8 Q k Z P o 3 6 X O F z I i x J Z Q p b m 8 l b E g V Z c a m U 7 E h e I s v / y X + S f 2 i 7 t 2 e 1 h q X R R p l O I B D O A Y P z q A B 1 9 A E H x g M 4 A l e 4 N U R z r P z 5 r z P W 0 t O M b M P v + B 8 f A O l T o 2 R < / l a t e x i t > z 0 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / I h k t E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 1 b b w V 2 p F Q W p A / E E L D r X 3 / o 6 T y Q = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S O I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9 b n D / I h k t E = < / l a t e x i t > z 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "b n D / O k k t I = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" j O 2 5 F i j P w O O / l y V m 1 C v R y T Q W w P c = \" > A A A B 9 3 i c b V B P S 8 M w H P 1 1 / p v z z 6 o e v Q S H 4 G m 0 I q i 3 o R e P E 6 w b b K W k a b q F p W l J U m E b + y R e P K h 4 9 a t 4 8 9 u Y b j 3 o 5 o O Q x 3 u / H 3 l 5 Y c a Z 0 o 7 z b V X W 1 j c 2 t 6 r b t Z 3 d v f 2 6 f X D 4 q N J c E u q R l K e y G 2 J F O R P U 0 0 x z 2 s 0 k x U n I a S c c 3",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "R Z + 5 4 l K x V L x o M c Z 9 R M 8 E C x m B G s j B X a 9 H 6 Y 8 U u P E X G g S u I H d c J r O H G i V u C V p Q I l 2 Y H / 1 o 5 T k C R W a c K x U z 3 U y 7 U + x 1 I x w O q v 1 c 0 U z T E Z 4 Q H u G C p x Q 5 U / n w W f o 1 C g R i l N p j t B o r v 7 e m O J E F d n M Z I L 1 U C 1 7 h f i f 1 8 t 1 f O V P m c h y T Q V Z P B T n H O k U F S 2 g i E l K N B 8 b g o l k J i s i Q y w x 0 a a r m i n B X f 7 y K v H O m 9 d N 9 / 6 i 0 b o p 2 6 j C M Z z A G b h w C S 2 4 g z Z 4 Q C C H Z 3 i F N 2 t i v V j v 1 s d i t G K V O 0 f w B 9 b n D / O k k t I = < / l a t e x i t > z K 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "< l a t e x i t s h a 1 _ b a s e 6 4 = \" Y / ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y V a 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "V I J S Y u F k z I X o A U Y Z Q T V 1 P N S C + R B M U B I 9 1 g f F X 4 3 Q c i F R X 8 T m c J 8 W I 0 5 D S i G G k j + f X G I B A s V F l s L v j o 5 z f H z s S v N + 2 W P Q V c J E 5 J m q B E x 6 9 / D U K B 0 5 h w j R l S q u / Y i f Z y J D X F j E x q g 1 S R B O E x G p K + o R z F R H n 5 N P w E H h o l h J G Q 5 n A N p + r v j R z F q s h n J m O k R 2 r e K 8 T / v H 6 q o 3 M v p z",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "x J N e F 4 9 l C U M q g F L J q A I Z U E a 5 Y Z g r C k J i v E I y Q R 1 q a v m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6 / M H z i 2 U a g = = < / l a t e x i t > z K < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" n U T d 1 V e U R H 8 K N G 6 Z F T Y 7 n k G S o j 4 = \" > A A A B + X i c b V B P S 8 M w H E 3 n v z n / d X r 0 E h y C p 9 G K o N 6 G X g Q v E 6 w b b K W k a b q F p U l J U m X W f R Q v H l S 8 + k 2 8 + W 1 M t x 5 0 8 0 H I 4 7 3 f j 7 y 8 M G V U a c f 5 t i p L y y u r a 9 X 1 2 s b m 1 v a O X d + 9 U y K T m H h Y M C G 7 I V K E U U 4 8 T T U j 3 V Q S l I S M d M L R Z e F 3 7 o l U V P B b P U 6 J n 6 A B p z H F S B s p s O v 9 U L B I j R N z w c c g v 5 4 E d s N p O l P A R e K W p A F K t A P 7 q x 8 J n C W E a 8 y Q U j 3 X S b W f I 6 k p Z m R S 6 2 e K p A i P 0 I D 0 D O U o I c r P p 9 E n 8 N A o E Y y F N I d r O F V / b + Q o U U U 6 M 5 k g P V T z X i H + 5 / U y H Z / 5 O e V p p g n H s 4 f i j E E t Y N E D j K g k W L O x I Q h L a r J C P E Q S Y W 3 a q p k S 3 P k v L x L v u H n e d G 9 O G q 2 L s o 0 q 2 A c H 4 A i 4 4 B S 0 w B V o A w 9 g 8 A C e w S t 4 s 5 6 s F + v d + p i N V q x y Z w / 8 g f X 5 A + i P k / g = < / l a t e x i t >",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "a t e x i t s h a 1 _ b a s e 6 4 = \" i d p B h g g 0 b I s k h z q e k z H H y d b t g w I = \" > A",
                "sec_num": null
            },
            {
                "text": "GRU GRU GRU y M < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 3 r a c + C w Y m Q S 8 s 7 v x V 1 I B 1 + 5 n 9 I = \" > A A A B 7 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 s S Q i q L e i F y 9 C B W M L b S i b 7 a Z d u t m E 3 Y l Q Q n + E F w 8 q X v 0 / 3 v w 3 b t s c t P X B w O O 9 G W b m h a k U B l 3 3 2 1 l a X l l d W y 9 t l D e 3 t n d 2 K 3 v 7 j y b J N O M + S 2 S i W y E 1 X A r F f R Q o e S v V n M a h 5 M 1 w e D P x m 0 9 c G 5 G o B x y l P I h p X 4 l I M I p W a o 6 6 + d 2 p N + 5 W q m 7 N n Y I s E q 8 g V S j Q 6 F a + O r 2 E Z T F X y C Q 1 p u 2 5 K Q Y 5 1 S i Y 5 O N y J z M 8 p W x I + 7 x t q a I x N 0 E + P X d M j q 3 S I 1 G i b S k k U / X 3 R E 5 j Y 0 Z x a D t j i g M z 7 0 3 E / 7 x 2 h t F l k A u V Z s g V m y 2 K M k k w I Z P f S U 9 o z l C O L K F M C 3 s r Y Q O q K U O b U N m G 4 M 2 / v E j 8 s 9 p V z b s / r 9 a v i z R K c A h H c A I e X E A d b q E B P j A Y w j O 8 w p u T O i / O u / M x a 1 1 y i p k D + A P n 8 w d E 4 4 8 P < / l a t e x i t > y < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" B 5 T i S o 8 2 P j I h i e 7 n A D a u 3 4 m R p N 8 = \" > A A A B 6 X i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G l t o Q 9 l s N + 3 S z S b s T o Q S + h O 8 e F D x 6 j / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m i T T j P s s k Y l u h 9 R w K R T 3 U a D k 7 V R z G o e S t 8 L R z d R v P X F t R K I e c J z y I K Y D J S L B K F r p f t x z e 9 W a W 3 d n I M v E K 0 g N C j R 7 1 a 9 u P 2 F Z z B U y S Y 3 p e G 6 K Q U 4 1 C i b 5 p N L N D E 8 p G 9 E B 7 1 i q a M x N k M 9 O n Z A T q / R J l G h b C s l M / T 2 R 0 9 i Y c R z a z p j i 0 C x 6 U / E / r 5 N h d B n k Q q U Z c s X m i 6 J M E k z I 9 G / S F 5 o z l G N L K N P C 3 k r Y k G r K 0 K Z T s S F 4 i y 8 v E / + s f l X 3 7 s 5 r j e s i j T I c w T G c g g c X 0 I B b a I I P D A b w D K / w 5 k j n x X l 3 P u a t J a e Y O Y Q / c D 5 / A H l 3 j X Q = < / l a t e x i t > y < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > y 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v D",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Q l p J w A v e 1 K U R F h Y Q W f K 1 8 h 1 0 0 t g = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T o Q Q + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g 1 Q c D j / d m m J k X p l I Y d N 0 v p 7 K y u r a + U d 2 s b W 3 v 7 O 7 V 9 w 8 e T J J p x n 2 W y E R 3 Q 2 q 4 F I r 7 K F D y b q o 5 j U P J O + H k Z u Z 3 H r k 2 I l H 3 m K c 8 i O l I i U g w i l b y 8 0 H h T Q f 1 h t t 0 5 y B / i V e S B p R o D + q f / W H C s p g r Z J I a 0 / P c F I O C a h R M 8 m m t n x m e U j a h I 9 6 z V N G Y m 6 C Y H z s l J 1 Y Z k i j R t h S S u f p z o q C x M X k c 2 s 6 Y 4 t g s e z P x P 6 + X Y The same neural topic model is also applied, but omitted here for simplicity of illustration. \"LT\" denotes a linear transformation.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "X Q Z F E K l G X L F F o u i T B J M y O x z M h S a M 5 S 5 J Z R p Y W 8 l b E w 1 Z W j z q d k Q v O W X / x L / r H n V 9 O 7 O G 6 3 r M o 0 q H M E x n I I H F 9 C C W 2 i D D w w E P M E L v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "using less conditional information while generating each word) Shen et al., 2017a) , or bridging the amortization gap (between the loglikelihood and the ELBO) using semi-amortized inference networks (Kim et al., 2018) . However, these methods mitigate the issue by weakening the conditional dependency on the decoder, which may fail to generate high-quality continuous sentences.",
                "cite_spans": [
                    {
                        "start": 63,
                        "end": 82,
                        "text": "Shen et al., 2017a)",
                        "ref_id": "BIBREF40"
                    },
                    {
                        "start": 199,
                        "end": 217,
                        "text": "(Kim et al., 2018)",
                        "ref_id": "BIBREF17"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "To overcome the two problems mentioned above, we propose a topic-guided variational autoencoder (TGVAE) model, permitting text generation with designated topic guidance. As illustrated in Figure 1(a) , TGVAE specifies a Gaussian mixture model (GMM) as the prior of the latent code, where each mixture component corresponds to a topic. The GMM is learnable based on a neural topic model -the mean and diagonal covariance of each mixture component is parameterized by the corresponding topic. Accordingly, the degree to which each component of the GMM is used to generate the latent code and the corresponding sentence is tied to the usage of the topics. In the inference phase, we initialize the latent code from a GMM generated via the encoder, and apply the invertiable Householder transformation (Bischof and Sun, 1994; Sun and Bischof, 1995) to derive the latent code with high flexibility and low complexity.",
                "cite_spans": [
                    {
                        "start": 798,
                        "end": 821,
                        "text": "(Bischof and Sun, 1994;",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 822,
                        "end": 844,
                        "text": "Sun and Bischof, 1995)",
                        "ref_id": "BIBREF43"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 188,
                        "end": 199,
                        "text": "Figure 1(a)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "As shown in Figure 1(b) , besides unconditional text generation, the proposed model can be extended for conditional text generation, i.e., abstractive text summarization (Nallapati et al., 2016) with an attention module. By injecting the topics learned by our model (semantic information), we are able to make better use of the source document and improve a sequence-to-sequence summarization model (Sutskever et al., 2014) .",
                "cite_spans": [
                    {
                        "start": 170,
                        "end": 194,
                        "text": "(Nallapati et al., 2016)",
                        "ref_id": "BIBREF31"
                    },
                    {
                        "start": 399,
                        "end": 423,
                        "text": "(Sutskever et al., 2014)",
                        "ref_id": "BIBREF44"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 12,
                        "end": 23,
                        "text": "Figure 1(b)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "We highlight the contributions of our model as follows: (i) A new Topic-Guided VAE (TGVAE) model is proposed for text generation with designated topic guidance. (ii) For the model inference, Householder flow is introduced to transform a relatively simple mixture distribution into an arbitrarily flexible approximate posterior, achieving powerful approximate posterior inference. (iii) Experiments for both unconditional and conditional text generation demonstrate the effectiveness of the proposed approach.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow",
                "sec_num": null
            },
            {
                "text": "The proposed TGVAE, as illustrated in Figure 1(a) , consists of two modules: a neural topic model (NTM) and a neural sequence model (NSM). The NTM aims to capture long-range semantic meaning across the document, while the NSM is designed to generate a sentence with designated topic guidance.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 38,
                        "end": 49,
                        "text": "Figure 1(a)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Model",
                "sec_num": "2"
            },
            {
                "text": "Let d \u2208 Z D + denote the bag-of-words representation of a document, with Z + denoting non-negative integers. D is the vocabulary size, and each element of d reflects a count of the number of times the corresponding word occurs in the document. Let a n represent the topic assignment for word w n . Following Miao et al. (2017) , a Gaussian random vector is passed through a softmax function to parameterize the multinomial document topic distributions. Specifically, the generative process of the NTM is \u03b8 \u223c N (0, I), t = g(\u03b8) ,",
                "cite_spans": [
                    {
                        "start": 308,
                        "end": 326,
                        "text": "Miao et al. (2017)",
                        "ref_id": "BIBREF28"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "(1) a n \u223c Discrete(t), w n \u223c Discrete(\u03b2 an ) , where N (0, I) is an isotropic Gaussian distribution, g(\u2022) is a transformation function that maps sample \u03b8 to the topic embedding t, defined here as g(\u03b8) = softmax(\u0174\u03b8 +b), where\u0174 andb are trainable parameters; \u03b2 an represents the distribution over words for topic a n ; n \u2208 [1, N d ], and N d is the number of words in the document. The marginal likelihood for document d is:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "p(d|\u03b2) = t p(t) n an p(wn|\u03b2 an )p(an|t)dt (2) = t p(t) n p(wn|\u03b2, t)dt = t p(t)p(d|\u03b2, t)dt = \u03b8 p(\u03b8)p(d|\u03b2, \u03b8)d\u03b8 .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "The second equation in (2) holds because we can marginalize out the sampled topic words a n by",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "p(wn|\u03b2, t) = an p(wn|\u03b2 an )p(an|t) = \u03b2t , (3) where \u03b2 = {\u03b2 i } T i=1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "are trainable parameters of the decoder; T is the number of topics and each \u03b2 i \u2208 R D is a topic distribution over words (all elements of \u03b2 i are nonnegative, and sum to one).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Topic Model",
                "sec_num": "2.1"
            },
            {
                "text": "Our neural sequence model for text generation is built upon the VAE proposed in Bowman et al. (2015). Specifically, a continuous latent code z is first generated from some prior distribution p(z), based on which the text sequence y is then generated from a conditional distribution p(y|z) parameterized by a neural network (often called the decoder). Since the model incorporates a latent variable z that modulates the entire generation of the sentence, it should be able to capture the highlevel source of variation in the data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "Topic-Guided Gaussian Mixture Prior The aforementioned intuition is hard to be captured by a standard VAE, simply imposing a Gaussian prior on top of z, since the semantic information associated with a document intrinsically contains different subgroups (such as topics, sentiment, etc.). In our model, we consider incorporating the topic information into latent variables. Our model assumes each z is drawn from a topic-dependent GMM, that is,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "p(z|\u03b2, t) = T i=1 t i N (\u00b5(\u03b2 i ), \u03c3 2 (\u03b2 i )) \u00b5(\u03b2 i ) = f \u00b5 (\u03b2 i ) \u03c3 2 (\u03b2 i ) = diag(exp (f \u03c3 (\u03b2 i ))) ,",
                        "eq_num": "(4)"
                    }
                ],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "where t i is the usage of topic i in a document and \u03b2 i is the i-th topic distribution over words. Both of them are inherited from the NTM discussed above. Both f \u00b5 (\u2022) and f \u03c3 (\u2022) are implemented as feedforward neural networks, with trainable parameters W \u00b5 and W \u03c3 , respectively. Compared with a normal GMM prior that sets each mixture component to be N (0, I), the proposed topic guided GMM prior provides semantic meaning for each mixture component, and hence makes the model more interpretable and controllable for text generation.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "Decoder The likelihood of a word sequence y = {y m } M m=1 conditioned on the latent code z is defined as:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "p(y|z) = p(y 1 |z) M m=2 p(y m |y 1:m\u22121 , z) = p(y 1 |z) M m=2 p(y m |h m ) ,",
                        "eq_num": "(5)"
                    }
                ],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "where the conditional probability of each word y m given all the previous words y 1:m\u22121 and the latent code z is defined through the hidden state h m :",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "h m = f (h m\u22121 , y m\u22121 , z),",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "where the function f (\u2022) is implemented as a Gated Recurrent Unit (GRU) cell (Cho et al., 2014) in our experiments.",
                "cite_spans": [
                    {
                        "start": 77,
                        "end": 95,
                        "text": "(Cho et al., 2014)",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Neural Sequence Model",
                "sec_num": "2.2"
            },
            {
                "text": "The proposed model (see Figure 1(a) ) takes the bagof-words as input and embeds a document into a topic vector. The topic vector is then used to reconstruct the bag-of-words input, and the learned topic distribution over words is used to model a topicdependent prior to generate a sentence in the VAE setup. Specifically, the joint marginal likelihood can be written as:",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 24,
                        "end": 35,
                        "text": "Figure 1(a)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Inference",
                "sec_num": "3"
            },
            {
                "text": "p(y, d|\u03b2) = \u03b8 z p(\u03b8)p(d|\u03b2, \u03b8) \u2022 p(z|\u03b2, \u03b8)p(y|z) d\u03b8dz . (6)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Inference",
                "sec_num": "3"
            },
            {
                "text": "Since direct optimization of (6) is intractable, autoencoding variational Bayes is employed (Kingma and Welling, 2013). Denote q(\u03b8|d) and q(z|y) as the variational distributions for \u03b8 and z, respectively. The variational objective function, also called the evidence lower bound (ELBO), is constructed as",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Inference",
                "sec_num": "3"
            },
            {
                "text": "L = E q(\u03b8|d) [log p(d|\u03b2, \u03b8)] \u2212 KL (q(\u03b8|d)||p(\u03b8)) neural topic model,L t + (7) E q(z|y) [log p(y|z)] \u2212 E q(\u03b8|d) [KL (q(z|y)||p(z|\u03b2, \u03b8))]",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Inference",
                "sec_num": "3"
            },
            {
                "text": "neural sequence model,Ls",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Inference",
                "sec_num": "3"
            },
            {
                "text": "By assuming",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": ".",
                "sec_num": null
            },
            {
                "text": "q(\u03b8|d) = N (\u03b8|g \u00b5 (d), diag(exp (g \u03c3 (d)))),",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": ".",
                "sec_num": null
            },
            {
                "text": "where both g \u00b5 (\u2022) and g \u03c3 (\u2022) are implemented as feed-forward neural networks, the reparameterization trick (Kingma and Welling, 2013) can be applied directly to build an unbiased and low-variance gradient estimator for the L t term in (7). Below, we discuss in detail how to approximate the L s term in (7) and infer an arbitrarily complex posterior for z. Note that z is henceforth represented as z K in preparation for the introduction of Householder flows.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": ".",
                "sec_num": null
            },
            {
                "text": "Householder flow Tomczak and Welling, 2016 ) is a volume-preserving normalizing flow (Rezende and Mohamed, 2015), capable of constructing an arbitrarily complex posterior q K (z K |y) from an initial random variable z 0 with distribution q 0 , by composing a sequence of invertible mappings, i.e., z",
                "cite_spans": [
                    {
                        "start": 17,
                        "end": 42,
                        "text": "Tomczak and Welling, 2016",
                        "ref_id": "BIBREF45"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "K = f K \u2022 \u2022 \u2022 \u2022 \u2022 f 2 \u2022 f 1 (z 0 ).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "By repeatedly applying the chain rule and using the property of Jacobians of invertible functions, q K (z K |y) is expressed as:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "log qK (zK |y) = log q0(z0|y) \u2212 K k=1 log det \u2202f k \u2202z k\u22121 ,",
                        "eq_num": "(8)"
                    }
                ],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "where | det \u2202f k \u2202z k\u22121 | is the absolute value of the Jacobian determinant. Therefore, the L s term in (7) may be rewritten as",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "E q 0 (z 0 |y) [log p(y|zK )] + K k=1 log det \u2202f k \u2202z k\u22121 \u2212E q(\u03b8|d) [KL(q0(z0|y)||p(zK |\u03b2, \u03b8))] .",
                        "eq_num": "(9)"
                    }
                ],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Here q 0 (z 0 |y) is also specified as a GMM, i.e., q 0 (z 0 |y) = T i=1 \u03c0 i (y)N (\u00b5 i (y), \u03c3 2 i (y)). As illustrated in Figure 1(a) , y is first represented as a hidden vector h, by encoding the text sequence with an RNN. Based on this, the mixture probabilities \u03c0, the means and diagonal covariances of all the mixture components are all produced by an encoder network, which is a linear layer with the input h. In (9), the first term can be considered as the reconstruction error, while the remaining two terms act as regularizers, the tractability of which is important for the whole framework.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 122,
                        "end": 133,
                        "text": "Figure 1(a)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "KL Divergence between two GMMs Since both the prior p(z K |\u03b2, \u03b8) and the initial density q 0 (z 0 |y) for the posterior are GMMs, the calculation of the third term in (9) requires the KL divergence between two GMMs. Though no closedform solutions exist, the KL divergence has an explicit upper bound (Dilokthanakul et al., 2016) , shown in Proposition 1.",
                "cite_spans": [
                    {
                        "start": 300,
                        "end": 328,
                        "text": "(Dilokthanakul et al., 2016)",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Proposition 1. For any two mixture densities p = n i=1 \u03c0 i g i andp = n i=1\u03c0 i\u011di , their KL divergence is upper-bounded by",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "KL (p||p) \u2264 KL (\u03c0||\u03c0) + n i=1 \u03c0iKL (gi||\u011di) , (10)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "where equality holds if and only if",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "\u03c0 i g i n i=1 \u03c0 i g i = \u03c0\u011d i n i=1\u03c0\u011d i . Proof. With the log-sum inequality KL (p||p) = i \u03c0igi log i \u03c0igi i\u03c0\u011d i \u2264 i \u03c0igi log \u03c0ig\u00ee \u03c0\u011di = i \u03c0i log \u03c0\u00ee \u03c0 + i \u03c0i gi log g\u00ee gi = KL(\u03c0||\u03c0) + i \u03c0iKL(gi||\u011di) .",
                        "eq_num": "(11)"
                    }
                ],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Since the KL divergence between two Gaussian distributions has a closed-form expression, the upper bound of the KL divergence between two GMMs can be readily calculated. Accordingly, the third term in (9) is upper bounded as",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "UKL = E q(\u03b8|d) KL (\u03c0(y)||t) (12) + T i=1 \u03c0i(y)KL N (\u00b5 i (y), \u03c3 2 i (y)||N (\u00b5(\u03b2 i ), \u03c3 2 (\u03b2 i )) ,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "where the expectation E q(\u03b8|d) [\u2022] can be approximated by a sample from q(\u03b8|d).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Householder Flow Householder flow (Tomczak and Welling, 2016) is a series of Householder transformations, defined as follows. For a given vector z k\u22121 , the reflection hyperplane can be defined by a Householder vector v t that is orthogonal to the hyperplane. The reflection of this point about the hyperplane is",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "z k = I \u2212 2 v k v T k ||v k || 2 z k\u22121 = H k z k\u22121 ,",
                        "eq_num": "(13)"
                    }
                ],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "where",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "H k = I \u2212 2 v k v T k ||v k || 2",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "is called the Householder matrix. An important property of the Householder matrix is that the absolute value of the Jacobian determinant is equal to 1, therefore K k=1 log det \u2202f k \u2202z k\u22121 = K k=1 log | det H k | = 0, significantly simplifying the computation of the lower bound in (9). For k = 1, . . . , K, the vector v k is produced by a linear layer with the input v k\u22121 , where v 0 = h is the last hidden vector of the encoder RNN that encodes the sentence y.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Finally, by combining (7), (9) and (12), the ELBO can be rewritten as 143.2 Extension to text summarization When extending our model to text summarization, we are interested in modeling p(y, d|x), where (x, y) denotes the documentsummary pair, and d denotes the bag-ofwords of the input document. The marginal likelihood can be written as p(y, d|x) = \u03b8 z p(\u03b8)p(d|\u03b2, \u03b8)p(z|\u03b2, \u03b8)p(y|x, z) d\u03b8dz. Assume the approximate posterior of z is only dependent on x, i.e., q(z|x) is proposed as the variational distribution for z. The ELBO is then constructed as",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "L \u2265 L t + E q 0 (z 0 |y) [log p(y|z K )] \u2212 U KL .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "L =L t + E q(z|x) [log p(y|x, z)] \u2212 E q(\u03b8|d) [KL (q(z|x)||p(z|\u03b2, \u03b8))] , (15)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "where L t is the same as used in (7). The main difference when compared with unconditional text generation lies in the usage of p(y|x, z) and q(z|x), illustrated in Figure 1(b) . The generation of y given x is not only dependent on a standard Seq2Seq model with attention (Nallapati et al., 2016) , but also affected by z (i.e., z K ), which provides the high-level topic guidance.",
                "cite_spans": [
                    {
                        "start": 272,
                        "end": 296,
                        "text": "(Nallapati et al., 2016)",
                        "ref_id": "BIBREF31"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 165,
                        "end": 176,
                        "text": "Figure 1(b)",
                        "ref_id": "FIGREF10"
                    }
                ],
                "eq_spans": [],
                "section": "Householder Flow for Approximate Posterior",
                "sec_num": "3.1"
            },
            {
                "text": "Redundancy in inferred topics is a common issue existing in general topic models. In order to address this, it is straightforward to regularize the row-wise distance between paired topics to diversify the topics. Following Xie et al. (2015) ; Miao et al. (2017) , we apply a topic diversity regularization while carrying out the inference.",
                "cite_spans": [
                    {
                        "start": 223,
                        "end": 240,
                        "text": "Xie et al. (2015)",
                        "ref_id": "BIBREF49"
                    },
                    {
                        "start": 243,
                        "end": 261,
                        "text": "Miao et al. (2017)",
                        "ref_id": "BIBREF28"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Diversity Regularizer for NTM",
                "sec_num": "3.3"
            },
            {
                "text": "Specifically, the distance between a pair of topics is measured by their cosine distance a(\u03b2 i , \u03b2 j ) = arccos",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Diversity Regularizer for NTM",
                "sec_num": "3.3"
            },
            {
                "text": "|\u03b2 i \u2022\u03b2 j | | \u03b2 i | 2 | \u03b2 j | 2 .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Diversity Regularizer for NTM",
                "sec_num": "3.3"
            },
            {
                "text": "The mean angle of all pairs of T topics is \u03c6 = 1 T 2 i j a(\u03b2 i , \u03b2 j ), and the variance is \u03bd = 1 T 2 i j (a(\u03b2 i , \u03b2 j ) \u2212 \u03c6) 2 . Finally, the topic-diversity regularization is defined as R = \u03c6 \u2212 \u03bd.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Diversity Regularizer for NTM",
                "sec_num": "3.3"
            },
            {
                "text": "The VAE was proposed by Kingma and Welling (2013) , and since then, it has been applied successfully in a variety of applications (Gregor et al., 2015; Kingma et al., 2014; Chen et al., 2017; Wang et al., 2018b; . Focusing on text generation, the methods in Miao et al. (2017 ; Srivastava and Sutton (2017) represent texts as bag-of-words, and Bowman et al. 2015proposed the usage of an RNN as the encoder and decoder, and found some negative results. In order to improve the performance, different convolutional designs (Semeniuta et al., 2017; Shen et al., 2017a; have been proposed. A VAE variant was further developed in to control the sentiment and tense of generated sentences. Additionally, the VAE has also been considered for conditional text generation tasks, including machine translation (Zhang et al., 2016) , image captioning (Pu et al., 2016) , dialogue generation (Serban et al., 2017; Zhao et al., 2017 ) and text summarization . In particular, distinct from the above works, we propose the usage of a topic-dependent prior to explicitly incorporate topic guidance into the text-generation framework.",
                "cite_spans": [
                    {
                        "start": 24,
                        "end": 49,
                        "text": "Kingma and Welling (2013)",
                        "ref_id": "BIBREF19"
                    },
                    {
                        "start": 130,
                        "end": 151,
                        "text": "(Gregor et al., 2015;",
                        "ref_id": "BIBREF13"
                    },
                    {
                        "start": 152,
                        "end": 172,
                        "text": "Kingma et al., 2014;",
                        "ref_id": "BIBREF18"
                    },
                    {
                        "start": 173,
                        "end": 191,
                        "text": "Chen et al., 2017;",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 192,
                        "end": 211,
                        "text": "Wang et al., 2018b;",
                        "ref_id": "BIBREF48"
                    },
                    {
                        "start": 258,
                        "end": 275,
                        "text": "Miao et al. (2017",
                        "ref_id": "BIBREF28"
                    },
                    {
                        "start": 278,
                        "end": 306,
                        "text": "Srivastava and Sutton (2017)",
                        "ref_id": "BIBREF42"
                    },
                    {
                        "start": 521,
                        "end": 545,
                        "text": "(Semeniuta et al., 2017;",
                        "ref_id": "BIBREF37"
                    },
                    {
                        "start": 546,
                        "end": 565,
                        "text": "Shen et al., 2017a;",
                        "ref_id": "BIBREF40"
                    },
                    {
                        "start": 800,
                        "end": 820,
                        "text": "(Zhang et al., 2016)",
                        "ref_id": "BIBREF53"
                    },
                    {
                        "start": 840,
                        "end": 857,
                        "text": "(Pu et al., 2016)",
                        "ref_id": "BIBREF32"
                    },
                    {
                        "start": 880,
                        "end": 901,
                        "text": "(Serban et al., 2017;",
                        "ref_id": "BIBREF38"
                    },
                    {
                        "start": 902,
                        "end": 919,
                        "text": "Zhao et al., 2017",
                        "ref_id": "BIBREF57"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "4"
            },
            {
                "text": "The idea of using learned topics to improve NLP tasks has been explored previously, including methods combining topic and neural language models (Ahn et al., 2016; Dieng et al., 2016; Lau et al., 2017; Mikolov and Zweig, 2012; , as well as leveraging topic and word embeddings (Liu et al., 2015; Xu et al., 2018) . Distinct from them, we propose the use of topics to guide the prior of a VAE, rather than only the language model (i.e., the decoder in a VAE setup). This provides more flexibility in text modeling and also the ability to infer the posterior on latent codes, which could be useful for visualization and downstream tasks.",
                "cite_spans": [
                    {
                        "start": 145,
                        "end": 163,
                        "text": "(Ahn et al., 2016;",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 164,
                        "end": 183,
                        "text": "Dieng et al., 2016;",
                        "ref_id": "BIBREF10"
                    },
                    {
                        "start": 184,
                        "end": 201,
                        "text": "Lau et al., 2017;",
                        "ref_id": "BIBREF20"
                    },
                    {
                        "start": 202,
                        "end": 226,
                        "text": "Mikolov and Zweig, 2012;",
                        "ref_id": "BIBREF30"
                    },
                    {
                        "start": 277,
                        "end": 295,
                        "text": "(Liu et al., 2015;",
                        "ref_id": null
                    },
                    {
                        "start": 296,
                        "end": 312,
                        "text": "Xu et al., 2018)",
                        "ref_id": "BIBREF50"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "4"
            },
            {
                "text": "Neural abstractive summarization was pioneered in Rush et al. (2015) , and it was followed and extended by Chopra et al. (2016) . Currently the RNN-based encoder-decoder framework with attention (Nallapati et al., 2016; See et al., 2017) remains popular in this area. Attention models typ-ically work as a keyword detector, which is similar to topic modeling in spirit. This fact motivated us to extend our topic-guided VAE model to text summarization.",
                "cite_spans": [
                    {
                        "start": 50,
                        "end": 68,
                        "text": "Rush et al. (2015)",
                        "ref_id": null
                    },
                    {
                        "start": 107,
                        "end": 127,
                        "text": "Chopra et al. (2016)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 195,
                        "end": 219,
                        "text": "(Nallapati et al., 2016;",
                        "ref_id": "BIBREF31"
                    },
                    {
                        "start": 220,
                        "end": 237,
                        "text": "See et al., 2017)",
                        "ref_id": "BIBREF36"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "4"
            },
            {
                "text": "We evaluate our TGVAE on text generation and text summarization, and interpret its improvements both quantitatively and qualitatively.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Experiments",
                "sec_num": "5"
            },
            {
                "text": "Dataset We conduct experiments on three publicly available corpora: APNEWS, IMDB and BNC. 1 APNEWS 2 is a collection of Associated Press news articles from 2009 to 2016. IMDB is a set of movie reviews collected by Maas et al. 2011, and BNC (BNC Consortium, 2007) is the written portion of the British National Corpus, which contains excerpts from journals, books, letters, essays, memoranda, news and other types of text. For the three corpora, we tokenize the words and sentences, lowercase all word tokens, and filter out word tokens that occur less than 10 times. For the topic model, we remove stop words in the documents and exclude the top 0.1% most frequent words and also words that appear less than 100 documents. A summary statistics is provided in Table 1 . Evaluation We first compare the perplexity of our neural sequence model with a variety of baselines. Further, we evaluate BLEU scores on the generated sentences, noted as test-BLEU and self -BLEU. test-BLEU (higher is better) evaluates the quality of generated sentences using a group of real test-set sentences as the reference, and self -BLEU (lower is better) mainly measures the diversity of generated samples (Zhu et al., 2018) . Setup For the neural topic model (NTM), we consider a 2-layer feed-forward neural network to model q(\u03b8|d), with 256 hidden units in each layer; ReLU is used as the activation function. The hyperparameter \u03bb for the neural topic model diversity regularizer is fixed to 0.1 across all the experiments. All the sentences in the paragraph are used to obtain the bag-of-words presentation d. The maximum number of words in a paragraph is set to 300. For the neural sequence model (NSM), we use bidirectional-GRU as the encoder and a standard GRU as the decoder. The hidden state of our 1 These three datasets can be downloaded from https://github.com/jhlau/topically-driven-language-model.",
                "cite_spans": [
                    {
                        "start": 240,
                        "end": 262,
                        "text": "(BNC Consortium, 2007)",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 1183,
                        "end": 1201,
                        "text": "(Zhu et al., 2018)",
                        "ref_id": "BIBREF59"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 759,
                        "end": 766,
                        "text": "Table 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "2 https://www.ap.org/en-gb/ GRU is fixed to 600 across all the three corpora. For the input sequence, we fix the sequence length to 30. In order to avoid overfitting, dropout with a rate of 0.4 is used in each GRU layer.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "Baseline We test the proposed method with different numbers of topics (components in GMM) and different numbers of Householder flows (i.e., K), and compare it with six baselines: (i) a standard language model (LM); (ii) a standard variational RNN auto-encoder (VAE); (iii) a Gaussian priorbased VAE with Householder Flow (VAE+HF); (iv) a standard LSTM language model with LDA as additional feature (LDA+LSTM); (v) Topic-RNN (Dieng et al., 2016) , a joint learning framework which learns a topic model and a language model simultaneously; (vi) TDLM (Lau et al., 2017), a joint learning framework which learns a convolutional based topic model and a language model simultaneously.",
                "cite_spans": [
                    {
                        "start": 424,
                        "end": 444,
                        "text": "(Dieng et al., 2016)",
                        "ref_id": "BIBREF10"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "Results The results in Table 3 show that the models trained with a VAE and its Householder extension does not outperform a well-optimized language model, and the KL term tends to be annealed with the increase of K. In comparison, our TGVAE achieves a lower perplexity upper bound, with a relative larger U KL . We attribute the improvements to our topic guided GMM model design, which provides additional topical clustering information in the latent space; the Householder flow also boosts the posterior inference for our TGVAE. We also observe consistent improvements with the number of topics, which demonstrates the efficiency of our TGVAE.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 23,
                        "end": 30,
                        "text": "Table 3",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "To verify the generative power of our TGVAE, we generate samples from our topic-dependent prior and compare various methods on the BLEU scores in Table 2 . With the increase of topic numbers, our TGVAE yields consistently better self -BLEU and a boost over test-BLEU relative to standard VAE models. We also show a group of sampled sentences drawn from a portion of topics in Table 5 . Our TGVAE is able to generate diverse sentences under topic guidance. When generating sentences under a mixture of topics, we draw multiple samples from the GMM and take z as the averaged sample.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 146,
                        "end": 153,
                        "text": "Table 2",
                        "ref_id": null
                    },
                    {
                        "start": 376,
                        "end": 383,
                        "text": "Table 5",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "Though this paper focuses on generating coherent topic-specific sentences rather than the learned topics themselves, we also evaluate the topic coherence (Lau et al., 2017) to show the rationality of our joint learning framework. We compute topic coher- Table 6 : Example generated summaries on GIGAWORDS. D is the source article, R means the reference summary, Seq2seq represents the summary generated from the Seq2Seq model. finance crime disease stock politics auto sports law globalization terrorist Figure 2 : The t-SNE visualization of 1, 000 samples drawn from the learned topic-guided Gaussian mixture prior and they can be best viewed in color. have in text generation. Additionally, for the vocabulary, we count the frequency of words in both the source article the target summary, and maintain the top 30,000 tokens as the source article and target summary vocabulary. For the NTM, we further remove top 0.3% words and infrequent words to get a topic model vocabulary in size of 8000. For the NTM, we follow the same setup as our text generation. In the NSM, we keep using bidirectional-GRU as the encoder and a standard GRU as the decoder. The hidden state is fixed to 400. An attention mechanism (Bahdanau et al., 2015) is applied in our sequence-to-sequence model. Baseline We compare our method with the following alternatives: (i) a standard sequence-tosequence model with attention (Bahdanau et al., 2015 ) (Seq2Seq); (ii) a model similar to our TG-VAE, but without the usage of the topic-dependent prior and Householder flow (Var-Seq2Seq); and (iii) a model similar to our TGVAE, but without the usage of the topic dependent prior (Var-Seq2Seq- APNEWS   students  animals  murder  weather  mega  syria  lawsuit  album  airlines  zacks  education  dogs  first-degree  corecasters  lottery  iran  appeals  music  rail  cents  schools  zoo  shooting  winds  powerball militants  justices  film  transit  earnings  math  bear  sentenced  rain  gambling  afgan  constitutional  songs  bridge  revenue  teachers  wildlife  gunshot  snow  jackpot  korea  judge  comedy  airport  income   IMDB   war  children  epsiode  name  detective  ethic  action  horror  negative  japanese  aircraft  cinderella  season  crawford  holmes  porn  batman  horror  stupid  miike  president  musical  episode  stanwyck  poirot  unfunny  king  zombie  horrible  kurosawa  war  beatles  sandler  gable  christie  sex  chan  werewolf  sucks  sadako HF).",
                "cite_spans": [
                    {
                        "start": 154,
                        "end": 172,
                        "text": "(Lau et al., 2017)",
                        "ref_id": "BIBREF20"
                    },
                    {
                        "start": 1209,
                        "end": 1232,
                        "text": "(Bahdanau et al., 2015)",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 1399,
                        "end": 1421,
                        "text": "(Bahdanau et al., 2015",
                        "ref_id": "BIBREF1"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 254,
                        "end": 261,
                        "text": "Table 6",
                        "ref_id": null
                    },
                    {
                        "start": 504,
                        "end": 512,
                        "text": "Figure 2",
                        "ref_id": null
                    },
                    {
                        "start": 1663,
                        "end": 2439,
                        "text": "APNEWS   students  animals  murder  weather  mega  syria  lawsuit  album  airlines  zacks  education  dogs  first-degree  corecasters  lottery  iran  appeals  music  rail  cents  schools  zoo  shooting  winds  powerball militants  justices  film  transit  earnings  math  bear  sentenced  rain  gambling  afgan  constitutional  songs  bridge  revenue  teachers  wildlife  gunshot  snow  jackpot  korea  judge  comedy  airport  income   IMDB   war  children  epsiode  name  detective  ethic  action  horror  negative  japanese  aircraft  cinderella  season  crawford  holmes  porn  batman  horror  stupid  miike  president  musical  episode  stanwyck  poirot  unfunny  king  zombie  horrible  kurosawa  war  beatles  sandler  gable  christie  sex  chan  werewolf  sucks  sadako",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Text Generation",
                "sec_num": "5.1"
            },
            {
                "text": "Results The results in Table 7 show that our TG-VAE achieves better performance than a variety of strong baseline methods on both GIGAWORDS and DUC-2004, demonstrating the practical value of our model. It is worthwhile to note that recently several much more complex CNN/RNN architectures have been proposed for abstract text summarization, such as SEASS (Zhou et al., 2017) , ConvS2S (Gehring et al., 2017) , and Reinforced-ConvS2S (Wang et al., 2018a) . In this work, we focus on a relatively simple RNN architecture for fair comparison. In such a way, we are able to conclude that the improvements on the results are mainly from our topic-guided text generation strategy. As can be seen, though the Var-Seq2Seq model achieves comparable performance with the standard Seq2Seq model, the usage of Householder flow for more flexible posterior inference boosts the performance. Additionally, by combining the proposed topic-dependent prior and Householder flow, we yield further performance improvements, demonstrating the importance of topic guidance for text summarization. To demonstrate the readability and diversity of the generated summaries, we present typical examples in Table 6 . The words in blue are the topic words that appear in the source article but do not exist in the reference, while the words in red are neither in the reference nor in the source article. When the topic information is provided, our model is able to generate semantically-meaningful words which may not even exist in the reference summaries and the source articles. Additionally, with our topic-guided model, we can always generate a summary with meaningful initial words. These phenomena imply that our model supplies more insightful semantic information to improve the quality of generated summaries.",
                "cite_spans": [
                    {
                        "start": 355,
                        "end": 374,
                        "text": "(Zhou et al., 2017)",
                        "ref_id": "BIBREF58"
                    },
                    {
                        "start": 385,
                        "end": 407,
                        "text": "(Gehring et al., 2017)",
                        "ref_id": "BIBREF12"
                    },
                    {
                        "start": 433,
                        "end": 453,
                        "text": "(Wang et al., 2018a)",
                        "ref_id": "BIBREF46"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 23,
                        "end": 30,
                        "text": "Table 7",
                        "ref_id": "TABREF1"
                    },
                    {
                        "start": 1179,
                        "end": 1186,
                        "text": "Table 6",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "GIGAWORDS",
                "sec_num": null
            },
            {
                "text": "Finally, to demonstrate that our TGVAE learns interpretable topic-dependent GMM priors, we draw multiple samples from each mixture component and visualize them with t-SNE (Maaten and Hinton, 2008) . As can be seen from Figure 2 , we have learned a group of separable topic-dependent components. Each component is clustered and also maintains semantic meaning in the latent space, e.g., the clusters corresponding to the topic \"stock\" and \"finance\" are close to each other, while the clusters for \"finance\" and \"disease\" are far away from each other. Additionally, to understand the topic model we have learned, we provide the top 5 words for 10 randomly chosen topics on each dataset (the boldface word is the topic name summarized by us), as shown in Table 8 .",
                "cite_spans": [
                    {
                        "start": 171,
                        "end": 196,
                        "text": "(Maaten and Hinton, 2008)",
                        "ref_id": "BIBREF26"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 219,
                        "end": 227,
                        "text": "Figure 2",
                        "ref_id": null
                    },
                    {
                        "start": 752,
                        "end": 759,
                        "text": "Table 8",
                        "ref_id": "TABREF4"
                    }
                ],
                "eq_spans": [],
                "section": "GIGAWORDS",
                "sec_num": null
            },
            {
                "text": "A novel text generator is developed, combining a VAE-based neural sequence model with a neural topic model. The model is an extension of conditional VAEs in the framework of unsupervised learning, in which the topics are extracted from the data with clustering structure rather than predefined labels. An effective inference method based on Householder flow is designed to encourage the complexity and the diversity of the learned topics. Experimental results are encouraging, across multiple NLP tasks.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "6"
            },
            {
                "text": "http://duc.nist.gov/duc2004",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [
            {
                "text": "Vocabulary Training Development Testing LM TM # Docs # Sents # Tokens # Docs # Sents # Tokens # Docs # Sents # Tokens APNEWS 32, 400 7 ence using normalized PMI (NPMI). In practice, we average topic coherence over the top 5/10/15/20 topic words. To aggregate topic coherence score, we further average the coherence scores over topics. Results are summarized in Table 4 .,",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 361,
                        "end": 368,
                        "text": "Table 4",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Dataset",
                "sec_num": null
            },
            {
                "text": "We further test our model for text summarization on two popular datasets. First, we follow the same setup as in Rush et al. (2015) input-summary pair consists of the first sentence and the headline of the source articles. We also evaluate various models on the DUC-2004 test set 4 , which has 500 news articles. Different from GIGAWORDS, each article in DUC-2004 is paired with four expert-generated reference summaries. The length of each summary is limited to 75 bytes. Evaluation We evaluate the performance of our model with the ROUGE score (Lin, 2004) , which counts the number of overlapping content between the generated summaries and the reference summaries, e.g., overlapped n-grams. Following practice, we use F-measures of ROUGE-1 (RF-1), ROUGE-2 (RF-2) and ROUGE-L (RF-L) for GI-GAWORDS and Recall measures of ROUGE-1 (RR-1), ROUGE-2 (RR-2) and ROUGE-L (RR-L) for DUC-2004. Setup We have a similar data tokenization as we Data Topic Sentences",
                "cite_spans": [
                    {
                        "start": 112,
                        "end": 130,
                        "text": "Rush et al. (2015)",
                        "ref_id": null
                    },
                    {
                        "start": 545,
                        "end": 556,
                        "text": "(Lin, 2004)",
                        "ref_id": "BIBREF23"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Text Summarization Dataset",
                "sec_num": "5.2"
            },
            {
                "text": "\u2022 the commission has approved a bill that would make state funding available for the city 's new school . animal\u2022the feline did n't survive fence hangars at the lake .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "APNEWS education",
                "sec_num": null
            },
            {
                "text": "\u2022 the jury found the defense was not a <unk> , <unk> 's ruling and that the state 's highest court has been convicted of first-degree murder . weather\u2022 forecasters say they 're still trying to see the national weather service watch for the latest forecast for friday evening . lottory\u2022 she hopes the jackpot now exceeds $ 9 million .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "crime",
                "sec_num": null
            },
            {
                "text": "\u2022 an alabama law professor thomas said monday that the state's open court claims it takes an emotional matter about issuing child molesters based on religion. animal+medicine\u2022 the study says the animal welfare department and others are not sure to make similar cases to the virus in the zoo.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "education+law",
                "sec_num": null
            },
            {
                "text": "\u2022 after watching the movie , there is a great documentary about the war in the years of the israeli war . children\u2022 the entire animation was great at times as to the readings of disney favorites . epsiode\u2022 the show would have warranted for 25 episodes and it does help immediately . name\u2022 she steals the other part where norma 's <unk> husband ( crawford ) ( as at his part , sh*t for the road ) . detective\u2022 holmes shouted just to be as much as the movie 's last scene where there were <unk> pills to nab the <unk> . horror + negative\u2022 the movie about a zombie is the worst movie i have ever seen. detective + children\u2022 my favorite childhood is that rochester takes the character in jane's way, playing the one with hamlet.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "IMDB war",
                "sec_num": null
            },
            {
                "text": "\u2022 here mistaking ' causes ' drugs as the problem although both economically ill patients arising from a local job will be in traumatic dangers .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "BNC medical",
                "sec_num": null
            },
            {
                "text": "\u2022 he says the sale is given to five students ' award off : out at a laboratory after the three watts of the hours travelling in and chairman store the bank of the <unk> sutcliffe . religion\u2022 schoolchildren will either go or back to church in his place every year in the savoy . entertainment\u2022 100 company and special lace with <unk> garland for tea our garden was filmed after a ceremony IT\u2022 ibm also has shut all the big macs in the 60mhz ncube , represent on the acquisition and mips unix . environment + crime\u2022 the earth's environmental protection agency said that the government was still being shut down by the police. education+entertainment \u2022 the school is 55 and hosts one of a musician's theme charities festival. Table 5 : Generated sentences from given topics.Sample of Summaries D: a court here thursday sentenced a ##-year-old man to ## years in jail after he admitted pummelling his baby son to death to silence him while watching television . R: man who killed baby to hear television better gets ## years. Seq2Seq: man sentenced to ## years after the son 's death Ours: a court sentenced a man ## years in jail D: european stock markets advanced strongly thursday on some bargain-hunting and gains by wall street and japanese shares ahead of an expected hike in us interest rates , dealers said R: european stocks bounce back UNK UNK with closing levels Seq2Seq: european stocks advance ahead of us interest rate hike Ours: european stocks rise on bargain-hunting, dealer said friday D: the democratic people 's republic of korea whitewashed south korea in the women 's team semi-finals at the world table tennis championships here on sunday R: dpr korea sails into women 's team final Seq2Seq: dpr korea whitewash south korea in women 's team final Ours: dpr korea beat south korea in table tennis worlds",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 723,
                        "end": 730,
                        "text": "Table 5",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "education",
                "sec_num": null
            }
        ],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "A neural knowledge language model",
                "authors": [
                    {
                        "first": "Heeyoul",
                        "middle": [],
                        "last": "Sungjin Ahn",
                        "suffix": ""
                    },
                    {
                        "first": "Tanel",
                        "middle": [],
                        "last": "Choi",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "P\u00e4rnamaa",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1608.00318"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Sungjin Ahn, Heeyoul Choi, Tanel P\u00e4rnamaa, and Yoshua Bengio. 2016. A neural knowledge language model. arXiv preprint arXiv:1608.00318.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Neural machine translation by jointly learning to align and translate",
                "authors": [
                    {
                        "first": "Dzmitry",
                        "middle": [],
                        "last": "Bahdanau",
                        "suffix": ""
                    },
                    {
                        "first": "Kyunghyun",
                        "middle": [],
                        "last": "Cho",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "ICLR",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Ben- gio. 2015. Neural machine translation by jointly learning to align and translate. In ICLR.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "On orthogonal block elimination",
                "authors": [
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Christian",
                        "suffix": ""
                    },
                    {
                        "first": "Xiaobai",
                        "middle": [],
                        "last": "Bischof",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Sun",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Christian H Bischof and Xiaobai Sun. 1994. On or- thogonal block elimination. Preprint MCS-P450- 0794, Mathematics and Computer Science Division, Argonne National Laboratory.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Latent dirichlet allocation. JMLR",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "David",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Blei",
                        "suffix": ""
                    },
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Andrew",
                        "suffix": ""
                    },
                    {
                        "first": "Michael I Jordan",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "David M Blei, Andrew Y Ng, and Michael I Jordan. 2003. Latent dirichlet allocation. JMLR.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "The British National Corpus, version 3 (BNC XML Edition)",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "Bnc Bnc Consortium",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "BNC BNC Consortium. 2007. The British National Corpus, version 3 (BNC XML Edition). Distributed by Bodleian Libraries, University of Oxford, on be- half of the BNC Consortium.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Generating sentences from a continuous space",
                "authors": [
                    {
                        "first": "Luke",
                        "middle": [],
                        "last": "Samuel R Bowman",
                        "suffix": ""
                    },
                    {
                        "first": "Oriol",
                        "middle": [],
                        "last": "Vilnis",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Vinyals",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Andrew",
                        "suffix": ""
                    },
                    {
                        "first": "Rafal",
                        "middle": [],
                        "last": "Dai",
                        "suffix": ""
                    },
                    {
                        "first": "Samy",
                        "middle": [],
                        "last": "Jozefowicz",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1511.06349"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Samuel R Bowman, Luke Vilnis, Oriol Vinyals, An- drew M Dai, Rafal Jozefowicz, and Samy Ben- gio. 2015. Generating sentences from a continuous space. arXiv preprint arXiv:1511.06349.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Continuous-time flows for efficient inference and density estimation",
                "authors": [
                    {
                        "first": "Changyou",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Chunyuan",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Liqun",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Yunchen",
                        "middle": [],
                        "last": "Pu",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1709.01179"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Changyou Chen, Chunyuan Li, Liqun Chen, Wen- lin Wang, Yunchen Pu, and Lawrence Carin. 2017. Continuous-time flows for efficient in- ference and density estimation. arXiv preprint arXiv:1709.01179.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Adversarial text generation via featuremover's distance",
                "authors": [
                    {
                        "first": "Liqun",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Shuyang",
                        "middle": [],
                        "last": "Dai",
                        "suffix": ""
                    },
                    {
                        "first": "Chenyang",
                        "middle": [],
                        "last": "Tao",
                        "suffix": ""
                    },
                    {
                        "first": "Haichao",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Zhe",
                        "middle": [],
                        "last": "Gan",
                        "suffix": ""
                    },
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Yizhe",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Guoyin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Ruiyi",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "NeurIPS",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Liqun Chen, Shuyang Dai, Chenyang Tao, Haichao Zhang, Zhe Gan, Dinghan Shen, Yizhe Zhang, Guoyin Wang, Ruiyi Zhang, and Lawrence Carin. 2018. Adversarial text generation via feature- mover's distance. In NeurIPS.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Learning phrase representations using RNN encoder-decoder for statistical machine translation",
                "authors": [
                    {
                        "first": "Kyunghyun",
                        "middle": [],
                        "last": "Cho",
                        "suffix": ""
                    },
                    {
                        "first": "Bart",
                        "middle": [],
                        "last": "Van Merri\u00ebnboer",
                        "suffix": ""
                    },
                    {
                        "first": "Caglar",
                        "middle": [],
                        "last": "Gulcehre",
                        "suffix": ""
                    },
                    {
                        "first": "Dzmitry",
                        "middle": [],
                        "last": "Bahdanau",
                        "suffix": ""
                    },
                    {
                        "first": "Fethi",
                        "middle": [],
                        "last": "Bougares",
                        "suffix": ""
                    },
                    {
                        "first": "Holger",
                        "middle": [],
                        "last": "Schwenk",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1406.1078"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Kyunghyun Cho, Bart Van Merri\u00ebnboer, Caglar Gul- cehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Abstractive sentence summarization with attentive recurrent neural networks",
                "authors": [
                    {
                        "first": "Sumit",
                        "middle": [],
                        "last": "Chopra",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [],
                        "last": "Auli",
                        "suffix": ""
                    },
                    {
                        "first": "Alexander M",
                        "middle": [],
                        "last": "Rush",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Sumit Chopra, Michael Auli, and Alexander M Rush. 2016. Abstractive sentence summarization with at- tentive recurrent neural networks. In NAACL.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Topicrnn: A recurrent neural network with long-range semantic dependency",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Adji",
                        "suffix": ""
                    },
                    {
                        "first": "Chong",
                        "middle": [],
                        "last": "Dieng",
                        "suffix": ""
                    },
                    {
                        "first": "Jianfeng",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "John",
                        "middle": [],
                        "last": "Gao",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Paisley",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1611.01702"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Adji B Dieng, Chong Wang, Jianfeng Gao, and John Paisley. 2016. Topicrnn: A recurrent neural net- work with long-range semantic dependency. arXiv preprint arXiv:1611.01702.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Deep unsupervised clustering with Gaussian mixture variational autoencoders",
                "authors": [
                    {
                        "first": "Nat",
                        "middle": [],
                        "last": "Dilokthanakul",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [
                            "M"
                        ],
                        "last": "Pedro",
                        "suffix": ""
                    },
                    {
                        "first": "Marta",
                        "middle": [],
                        "last": "Mediano",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Garnelo",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [
                            "H"
                        ],
                        "last": "Matthew",
                        "suffix": ""
                    },
                    {
                        "first": "Hugh",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "Kai",
                        "middle": [],
                        "last": "Salimbeni",
                        "suffix": ""
                    },
                    {
                        "first": "Murray",
                        "middle": [],
                        "last": "Arulkumaran",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Shanahan",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1611.02648"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Nat Dilokthanakul, Pedro AM Mediano, Marta Garnelo, Matthew CH Lee, Hugh Salimbeni, Kai Arulkumaran, and Murray Shanahan. 2016. Deep unsupervised clustering with Gaussian mix- ture variational autoencoders. arXiv preprint arXiv:1611.02648.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "Convolutional sequence to sequence learning",
                "authors": [
                    {
                        "first": "Jonas",
                        "middle": [],
                        "last": "Gehring",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [],
                        "last": "Auli",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Grangier",
                        "suffix": ""
                    },
                    {
                        "first": "Denis",
                        "middle": [],
                        "last": "Yarats",
                        "suffix": ""
                    },
                    {
                        "first": "Yann N",
                        "middle": [],
                        "last": "Dauphin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1705.03122"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jonas Gehring, Michael Auli, David Grangier, De- nis Yarats, and Yann N Dauphin. 2017. Convolu- tional sequence to sequence learning. arXiv preprint arXiv:1705.03122.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "DRAW: A recurrent neural network for image generation",
                "authors": [
                    {
                        "first": "Karol",
                        "middle": [],
                        "last": "Gregor",
                        "suffix": ""
                    },
                    {
                        "first": "Ivo",
                        "middle": [],
                        "last": "Danihelka",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Graves",
                        "suffix": ""
                    },
                    {
                        "first": "Danilo",
                        "middle": [],
                        "last": "Jimenez Rezende",
                        "suffix": ""
                    },
                    {
                        "first": "Daan",
                        "middle": [],
                        "last": "Wierstra",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1502.04623"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Karol Gregor, Ivo Danihelka, Alex Graves, Danilo Jimenez Rezende, and Daan Wierstra. 2015. DRAW: A recurrent neural network for im- age generation. arXiv preprint arXiv:1502.04623.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Long text generation via adversarial training with leaked information",
                "authors": [
                    {
                        "first": "Jiaxian",
                        "middle": [],
                        "last": "Guo",
                        "suffix": ""
                    },
                    {
                        "first": "Sidi",
                        "middle": [],
                        "last": "Lu",
                        "suffix": ""
                    },
                    {
                        "first": "Han",
                        "middle": [],
                        "last": "Cai",
                        "suffix": ""
                    },
                    {
                        "first": "Weinan",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Yong",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    },
                    {
                        "first": "Jun",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1709.08624"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jiaxian Guo, Sidi Lu, Han Cai, Weinan Zhang, Yong Yu, and Jun Wang. 2017. Long text generation via adversarial training with leaked information. arXiv preprint arXiv:1709.08624.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Toward controlled generation of text",
                "authors": [
                    {
                        "first": "Zhiting",
                        "middle": [],
                        "last": "Hu",
                        "suffix": ""
                    },
                    {
                        "first": "Zichao",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    },
                    {
                        "first": "Xiaodan",
                        "middle": [],
                        "last": "Liang",
                        "suffix": ""
                    },
                    {
                        "first": "Ruslan",
                        "middle": [],
                        "last": "Salakhutdinov",
                        "suffix": ""
                    },
                    {
                        "first": "Eric",
                        "middle": [
                            "P"
                        ],
                        "last": "Xing",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "ICML",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Zhiting Hu, Zichao Yang, Xiaodan Liang, Ruslan Salakhutdinov, and Eric P Xing. 2017. Toward con- trolled generation of text. In ICML.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Variational deep embedding: An unsupervised and generative approach to clustering",
                "authors": [
                    {
                        "first": "Zhuxi",
                        "middle": [],
                        "last": "Jiang",
                        "suffix": ""
                    },
                    {
                        "first": "Yin",
                        "middle": [],
                        "last": "Zheng",
                        "suffix": ""
                    },
                    {
                        "first": "Huachun",
                        "middle": [],
                        "last": "Tan",
                        "suffix": ""
                    },
                    {
                        "first": "Bangsheng",
                        "middle": [],
                        "last": "Tang",
                        "suffix": ""
                    },
                    {
                        "first": "Hanning",
                        "middle": [],
                        "last": "Zhou",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1611.05148"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Zhuxi Jiang, Yin Zheng, Huachun Tan, Bangsheng Tang, and Hanning Zhou. 2016. Variational deep embedding: An unsupervised and gener- ative approach to clustering. arXiv preprint arXiv:1611.05148.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Semi-Amortized variational autoencoders",
                "authors": [
                    {
                        "first": "Yoon",
                        "middle": [],
                        "last": "Kim",
                        "suffix": ""
                    },
                    {
                        "first": "Sam",
                        "middle": [],
                        "last": "Wiseman",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Andrew",
                        "suffix": ""
                    },
                    {
                        "first": "David",
                        "middle": [],
                        "last": "Miller",
                        "suffix": ""
                    },
                    {
                        "first": "Alexander M",
                        "middle": [],
                        "last": "Sontag",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Rush",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1802.02550"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yoon Kim, Sam Wiseman, Andrew C Miller, David Sontag, and Alexander M Rush. 2018. Semi- Amortized variational autoencoders. arXiv preprint arXiv:1802.02550.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Semi-supervised learning with deep generative models",
                "authors": [
                    {
                        "first": "Shakir",
                        "middle": [],
                        "last": "Diederik P Kingma",
                        "suffix": ""
                    },
                    {
                        "first": "Danilo",
                        "middle": [],
                        "last": "Mohamed",
                        "suffix": ""
                    },
                    {
                        "first": "Max",
                        "middle": [],
                        "last": "Jimenez Rezende",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Welling",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "NIPS",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Diederik P Kingma, Shakir Mohamed, Danilo Jimenez Rezende, and Max Welling. 2014. Semi-supervised learning with deep generative models. In NIPS.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Autoencoding variational Bayes",
                "authors": [
                    {
                        "first": "P",
                        "middle": [],
                        "last": "Diederik",
                        "suffix": ""
                    },
                    {
                        "first": "Max",
                        "middle": [],
                        "last": "Kingma",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Welling",
                        "suffix": ""
                    }
                ],
                "year": 2013,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1312.6114"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Diederik P Kingma and Max Welling. 2013. Auto- encoding variational Bayes. arXiv preprint arXiv:1312.6114.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "Topically driven neural language model",
                "authors": [
                    {
                        "first": "Timothy",
                        "middle": [],
                        "last": "Jey Han Lau",
                        "suffix": ""
                    },
                    {
                        "first": "Trevor",
                        "middle": [],
                        "last": "Baldwin",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Cohn",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1704.08012"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jey Han Lau, Timothy Baldwin, and Trevor Cohn. 2017. Topically driven neural language model. arXiv preprint arXiv:1704.08012.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Adversarial learning for neural dialogue generation",
                "authors": [
                    {
                        "first": "Jiwei",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Will",
                        "middle": [],
                        "last": "Monroe",
                        "suffix": ""
                    },
                    {
                        "first": "Tianlin",
                        "middle": [],
                        "last": "Shi",
                        "suffix": ""
                    },
                    {
                        "first": "S\u00e9bastien",
                        "middle": [],
                        "last": "Jean",
                        "suffix": ""
                    },
                    {
                        "first": "Alan",
                        "middle": [],
                        "last": "Ritter",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Jurafsky",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1701.06547"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jiwei Li, Will Monroe, Tianlin Shi, S\u00e9bastien Jean, Alan Ritter, and Dan Jurafsky. 2017a. Adversar- ial learning for neural dialogue generation. arXiv preprint arXiv:1701.06547.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Deep recurrent generative decoder for abstractive text summarization",
                "authors": [
                    {
                        "first": "Piji",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Wai",
                        "middle": [],
                        "last": "Lam",
                        "suffix": ""
                    },
                    {
                        "first": "Lidong",
                        "middle": [],
                        "last": "Bing",
                        "suffix": ""
                    },
                    {
                        "first": "Zihao",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1708.00625"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Piji Li, Wai Lam, Lidong Bing, and Zihao Wang. 2017b. Deep recurrent generative decoder for abstractive text summarization. arXiv preprint arXiv:1708.00625.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "Rouge: A package for automatic evaluation of summaries. Text Summarization Branches Out",
                "authors": [
                    {
                        "first": "Chin-Yew",
                        "middle": [],
                        "last": "Lin",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chin-Yew Lin. 2004. Rouge: A package for auto- matic evaluation of summaries. Text Summarization Branches Out.",
                "links": null
            },
            "BIBREF24": {
                "ref_id": "b24",
                "title": "Tat-Seng Chua, and Maosong Sun. 2015. Topical word embeddings. In AAAI",
                "authors": [
                    {
                        "first": "Yang",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Zhiyuan",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yang Liu, Zhiyuan Liu, Tat-Seng Chua, and Maosong Sun. 2015. Topical word embeddings. In AAAI.",
                "links": null
            },
            "BIBREF25": {
                "ref_id": "b25",
                "title": "Learning word vectors for sentiment analysis",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Andrew",
                        "suffix": ""
                    },
                    {
                        "first": "Raymond",
                        "middle": [
                            "E"
                        ],
                        "last": "Maas",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Daly",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Peter",
                        "suffix": ""
                    },
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Pham",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Y",
                        "middle": [],
                        "last": "Andrew",
                        "suffix": ""
                    },
                    {
                        "first": "Christopher",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Potts",
                        "suffix": ""
                    }
                ],
                "year": 2011,
                "venue": "ACL",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Andrew L Maas, Raymond E Daly, Peter T Pham, Dan Huang, Andrew Y Ng, and Christopher Potts. 2011. Learning word vectors for sentiment analysis. In ACL.",
                "links": null
            },
            "BIBREF26": {
                "ref_id": "b26",
                "title": "Visualizing data using t-SNE. JMLR",
                "authors": [
                    {
                        "first": "Laurens",
                        "middle": [],
                        "last": "Van Der Maaten",
                        "suffix": ""
                    },
                    {
                        "first": "Geoffrey",
                        "middle": [],
                        "last": "Hinton",
                        "suffix": ""
                    }
                ],
                "year": 2008,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. JMLR.",
                "links": null
            },
            "BIBREF27": {
                "ref_id": "b27",
                "title": "Language as a latent variable: Discrete generative models for sentence compression",
                "authors": [
                    {
                        "first": "Yishu",
                        "middle": [],
                        "last": "Miao",
                        "suffix": ""
                    },
                    {
                        "first": "Phil",
                        "middle": [],
                        "last": "Blunsom",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1609.07317"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yishu Miao and Phil Blunsom. 2016. Language as a latent variable: Discrete generative mod- els for sentence compression. arXiv preprint arXiv:1609.07317.",
                "links": null
            },
            "BIBREF28": {
                "ref_id": "b28",
                "title": "Discovering discrete latent topics with neural variational inference",
                "authors": [
                    {
                        "first": "Yishu",
                        "middle": [],
                        "last": "Miao",
                        "suffix": ""
                    },
                    {
                        "first": "Edward",
                        "middle": [],
                        "last": "Grefenstette",
                        "suffix": ""
                    },
                    {
                        "first": "Phil",
                        "middle": [],
                        "last": "Blunsom",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1706.00359"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yishu Miao, Edward Grefenstette, and Phil Blun- som. 2017. Discovering discrete latent topics with neural variational inference. arXiv preprint arXiv:1706.00359.",
                "links": null
            },
            "BIBREF29": {
                "ref_id": "b29",
                "title": "Neural variational inference for text processing",
                "authors": [
                    {
                        "first": "Yishu",
                        "middle": [],
                        "last": "Miao",
                        "suffix": ""
                    },
                    {
                        "first": "Lei",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    },
                    {
                        "first": "Phil",
                        "middle": [],
                        "last": "Blunsom",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "ICML",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yishu Miao, Lei Yu, and Phil Blunsom. 2016. Neural variational inference for text processing. In ICML.",
                "links": null
            },
            "BIBREF30": {
                "ref_id": "b30",
                "title": "Context dependent recurrent neural network language model",
                "authors": [
                    {
                        "first": "Tomas",
                        "middle": [],
                        "last": "Mikolov",
                        "suffix": ""
                    },
                    {
                        "first": "Geoffrey",
                        "middle": [],
                        "last": "Zweig",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Tomas Mikolov and Geoffrey Zweig. 2012. Context dependent recurrent neural network language model. SLT.",
                "links": null
            },
            "BIBREF31": {
                "ref_id": "b31",
                "title": "Abstractive text summarization using sequence-to-sequence RNNs and beyond",
                "authors": [
                    {
                        "first": "Ramesh",
                        "middle": [],
                        "last": "Nallapati",
                        "suffix": ""
                    },
                    {
                        "first": "Bowen",
                        "middle": [],
                        "last": "Zhou",
                        "suffix": ""
                    },
                    {
                        "first": "Caglar",
                        "middle": [],
                        "last": "Gulcehre",
                        "suffix": ""
                    },
                    {
                        "first": "Bing",
                        "middle": [],
                        "last": "Xiang",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1602.06023"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Ramesh Nallapati, Bowen Zhou, Caglar Gulcehre, Bing Xiang, et al. 2016. Abstractive text summariza- tion using sequence-to-sequence RNNs and beyond. arXiv preprint arXiv:1602.06023.",
                "links": null
            },
            "BIBREF32": {
                "ref_id": "b32",
                "title": "Variational autoencoder for deep learning of images, labels and captions",
                "authors": [
                    {
                        "first": "Yunchen",
                        "middle": [],
                        "last": "Pu",
                        "suffix": ""
                    },
                    {
                        "first": "Zhe",
                        "middle": [],
                        "last": "Gan",
                        "suffix": ""
                    },
                    {
                        "first": "Ricardo",
                        "middle": [],
                        "last": "Henao",
                        "suffix": ""
                    },
                    {
                        "first": "Xin",
                        "middle": [],
                        "last": "Yuan",
                        "suffix": ""
                    },
                    {
                        "first": "Chunyuan",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Andrew",
                        "middle": [],
                        "last": "Stevens",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "NIPS",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yunchen Pu, Zhe Gan, Ricardo Henao, Xin Yuan, Chunyuan Li, Andrew Stevens, and Lawrence Carin. 2016. Variational autoencoder for deep learning of images, labels and captions. In NIPS.",
                "links": null
            },
            "BIBREF33": {
                "ref_id": "b33",
                "title": "Variational inference with normalizing flows",
                "authors": [
                    {
                        "first": "Danilo",
                        "middle": [],
                        "last": "Jimenez Rezende",
                        "suffix": ""
                    },
                    {
                        "first": "Shakir",
                        "middle": [],
                        "last": "Mohamed",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1505.05770"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Danilo Jimenez Rezende and Shakir Mohamed. 2015. Variational inference with normalizing flows. arXiv preprint arXiv:1505.05770.",
                "links": null
            },
            "BIBREF34": {
                "ref_id": "b34",
                "title": "Stochastic backpropagation and approximate inference in deep generative models",
                "authors": [
                    {
                        "first": "Danilo",
                        "middle": [],
                        "last": "Jimenez Rezende",
                        "suffix": ""
                    },
                    {
                        "first": "Shakir",
                        "middle": [],
                        "last": "Mohamed",
                        "suffix": ""
                    },
                    {
                        "first": "Daan",
                        "middle": [],
                        "last": "Wierstra",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1401.4082"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Danilo Jimenez Rezende, Shakir Mohamed, and Daan Wierstra. 2014. Stochastic backpropagation and approximate inference in deep generative models. arXiv preprint arXiv:1401.4082.",
                "links": null
            },
            "BIBREF35": {
                "ref_id": "b35",
                "title": "Sumit Chopra, and Jason Weston. 2015. A neural attention model for abstractive sentence summarization",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Alexander",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Rush",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1509.00685"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Alexander M Rush, Sumit Chopra, and Jason We- ston. 2015. A neural attention model for ab- stractive sentence summarization. arXiv preprint arXiv:1509.00685.",
                "links": null
            },
            "BIBREF36": {
                "ref_id": "b36",
                "title": "Get to the point: Summarization with pointer-generator networks",
                "authors": [
                    {
                        "first": "Abigail",
                        "middle": [],
                        "last": "See",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Peter",
                        "suffix": ""
                    },
                    {
                        "first": "Christopher D",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Manning",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1704.04368"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Abigail See, Peter J Liu, and Christopher D Man- ning. 2017. Get to the point: Summarization with pointer-generator networks. arXiv preprint arXiv:1704.04368.",
                "links": null
            },
            "BIBREF37": {
                "ref_id": "b37",
                "title": "A hybrid convolutional variational autoencoder for text generation",
                "authors": [
                    {
                        "first": "Stanislau",
                        "middle": [],
                        "last": "Semeniuta",
                        "suffix": ""
                    },
                    {
                        "first": "Aliaksei",
                        "middle": [],
                        "last": "Severyn",
                        "suffix": ""
                    },
                    {
                        "first": "Erhardt",
                        "middle": [],
                        "last": "Barth",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1702.02390"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Stanislau Semeniuta, Aliaksei Severyn, and Erhardt Barth. 2017. A hybrid convolutional variational autoencoder for text generation. arXiv preprint arXiv:1702.02390.",
                "links": null
            },
            "BIBREF38": {
                "ref_id": "b38",
                "title": "A hierarchical latent variable encoder-decoder model for generating dialogues",
                "authors": [
                    {
                        "first": "Iulian",
                        "middle": [],
                        "last": "Vlad Serban",
                        "suffix": ""
                    },
                    {
                        "first": "Alessandro",
                        "middle": [],
                        "last": "Sordoni",
                        "suffix": ""
                    },
                    {
                        "first": "Ryan",
                        "middle": [],
                        "last": "Lowe",
                        "suffix": ""
                    },
                    {
                        "first": "Laurent",
                        "middle": [],
                        "last": "Charlin",
                        "suffix": ""
                    },
                    {
                        "first": "Joelle",
                        "middle": [],
                        "last": "Pineau",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Aaron",
                        "suffix": ""
                    },
                    {
                        "first": "Yoshua",
                        "middle": [],
                        "last": "Courville",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Bengio",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "AAAI",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Iulian Vlad Serban, Alessandro Sordoni, Ryan Lowe, Laurent Charlin, Joelle Pineau, Aaron C Courville, and Yoshua Bengio. 2017. A hierarchical latent variable encoder-decoder model for generating dia- logues. In AAAI.",
                "links": null
            },
            "BIBREF39": {
                "ref_id": "b39",
                "title": "Nash: Toward end-to-end neural architecture for generative semantic hashing",
                "authors": [
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Qinliang",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Paidamoyo",
                        "middle": [],
                        "last": "Chapfuwa",
                        "suffix": ""
                    },
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Guoyin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    },
                    {
                        "first": "Ricardo",
                        "middle": [],
                        "last": "Henao",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1805.05361"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Dinghan Shen, Qinliang Su, Paidamoyo Chapfuwa, Wenlin Wang, Guoyin Wang, Lawrence Carin, and Ricardo Henao. 2018. Nash: Toward end-to-end neural architecture for generative semantic hashing. arXiv preprint arXiv:1805.05361.",
                "links": null
            },
            "BIBREF40": {
                "ref_id": "b40",
                "title": "Deconvolutional latent-variable model for text sequence matching",
                "authors": [
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Yizhe",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Ricardo",
                        "middle": [],
                        "last": "Henao",
                        "suffix": ""
                    },
                    {
                        "first": "Qinliang",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1709.07109"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Dinghan Shen, Yizhe Zhang, Ricardo Henao, Qinliang Su, and Lawrence Carin. 2017a. Deconvolutional latent-variable model for text sequence matching. arXiv preprint arXiv:1709.07109.",
                "links": null
            },
            "BIBREF41": {
                "ref_id": "b41",
                "title": "A conditional variational framework for dialog generation",
                "authors": [
                    {
                        "first": "Xiaoyu",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Hui",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Yanran",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Wenjie",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Shuzi",
                        "middle": [],
                        "last": "Niu",
                        "suffix": ""
                    },
                    {
                        "first": "Yang",
                        "middle": [],
                        "last": "Zhao",
                        "suffix": ""
                    },
                    {
                        "first": "Akiko",
                        "middle": [],
                        "last": "Aizawa",
                        "suffix": ""
                    },
                    {
                        "first": "Guoping",
                        "middle": [],
                        "last": "Long",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1705.00316"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Xiaoyu Shen, Hui Su, Yanran Li, Wenjie Li, Shuzi Niu, Yang Zhao, Akiko Aizawa, and Guoping Long. 2017b. A conditional variational framework for dia- log generation. arXiv preprint arXiv:1705.00316.",
                "links": null
            },
            "BIBREF42": {
                "ref_id": "b42",
                "title": "Autoencoding variational inference for topic models",
                "authors": [
                    {
                        "first": "Akash",
                        "middle": [],
                        "last": "Srivastava",
                        "suffix": ""
                    },
                    {
                        "first": "Charles",
                        "middle": [],
                        "last": "Sutton",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1703.01488"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Akash Srivastava and Charles Sutton. 2017. Autoen- coding variational inference for topic models. arXiv preprint arXiv:1703.01488.",
                "links": null
            },
            "BIBREF43": {
                "ref_id": "b43",
                "title": "A basiskernel representation of orthogonal matrices. SIAM journal on matrix analysis and applications",
                "authors": [
                    {
                        "first": "Xiaobai",
                        "middle": [],
                        "last": "Sun",
                        "suffix": ""
                    },
                    {
                        "first": "Christian",
                        "middle": [],
                        "last": "Bischof",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Xiaobai Sun and Christian Bischof. 1995. A basis- kernel representation of orthogonal matrices. SIAM journal on matrix analysis and applications.",
                "links": null
            },
            "BIBREF44": {
                "ref_id": "b44",
                "title": "Sequence to sequence learning with neural networks",
                "authors": [
                    {
                        "first": "Ilya",
                        "middle": [],
                        "last": "Sutskever",
                        "suffix": ""
                    },
                    {
                        "first": "Oriol",
                        "middle": [],
                        "last": "Vinyals",
                        "suffix": ""
                    },
                    {
                        "first": "Quoc V",
                        "middle": [],
                        "last": "Le",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ilya Sutskever, Oriol Vinyals, and Quoc V Le. 2014. Sequence to sequence learning with neural networks. In NIPS.",
                "links": null
            },
            "BIBREF45": {
                "ref_id": "b45",
                "title": "Improving variational auto-encoders using householder flow",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Jakub",
                        "suffix": ""
                    },
                    {
                        "first": "Max",
                        "middle": [],
                        "last": "Tomczak",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Welling",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1611.09630"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jakub M Tomczak and Max Welling. 2016. Improving variational auto-encoders using householder flow. arXiv preprint arXiv:1611.09630.",
                "links": null
            },
            "BIBREF46": {
                "ref_id": "b46",
                "title": "A reinforced topicaware convolutional sequence-to-sequence model for abstractive text summarization",
                "authors": [
                    {
                        "first": "Li",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Junlin",
                        "middle": [],
                        "last": "Yao",
                        "suffix": ""
                    },
                    {
                        "first": "Yunzhe",
                        "middle": [],
                        "last": "Tao",
                        "suffix": ""
                    },
                    {
                        "first": "Li",
                        "middle": [],
                        "last": "Zhong",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Qiang",
                        "middle": [],
                        "last": "Du",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1805.03616"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Li Wang, Junlin Yao, Yunzhe Tao, Li Zhong, Wei Liu, and Qiang Du. 2018a. A reinforced topic- aware convolutional sequence-to-sequence model for abstractive text summarization. arXiv preprint arXiv:1805.03616.",
                "links": null
            },
            "BIBREF47": {
                "ref_id": "b47",
                "title": "Topic compositional neural language model",
                "authors": [
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Zhe",
                        "middle": [],
                        "last": "Gan",
                        "suffix": ""
                    },
                    {
                        "first": "Wenqi",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Jiaji",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Ping",
                        "suffix": ""
                    },
                    {
                        "first": "Sanjeev",
                        "middle": [],
                        "last": "Satheesh",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1712.09783"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Wenlin Wang, Zhe Gan, Wenqi Wang, Dinghan Shen, Jiaji Huang, Wei Ping, Sanjeev Satheesh, and Lawrence Carin. 2017. Topic compositional neural language model. arXiv preprint arXiv:1712.09783.",
                "links": null
            },
            "BIBREF48": {
                "ref_id": "b48",
                "title": "Zero-shot learning via class-conditioned deep generative models",
                "authors": [
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Yunchen",
                        "middle": [],
                        "last": "Pu",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Vinay Kumar",
                        "suffix": ""
                    },
                    {
                        "first": "Kai",
                        "middle": [],
                        "last": "Verma",
                        "suffix": ""
                    },
                    {
                        "first": "Yizhe",
                        "middle": [],
                        "last": "Fan",
                        "suffix": ""
                    },
                    {
                        "first": "Changyou",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Piyush",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Rai",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "AAAI",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Wenlin Wang, Yunchen Pu, Vinay Kumar Verma, Kai Fan, Yizhe Zhang, Changyou Chen, Piyush Rai, and Lawrence Carin. 2018b. Zero-shot learning via class-conditioned deep generative models. In AAAI.",
                "links": null
            },
            "BIBREF49": {
                "ref_id": "b49",
                "title": "Diversifying restricted Boltzmann machine for document modeling",
                "authors": [
                    {
                        "first": "Pengtao",
                        "middle": [],
                        "last": "Xie",
                        "suffix": ""
                    },
                    {
                        "first": "Yuntian",
                        "middle": [],
                        "last": "Deng",
                        "suffix": ""
                    },
                    {
                        "first": "Eric",
                        "middle": [],
                        "last": "Xing",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "KDD",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Pengtao Xie, Yuntian Deng, and Eric Xing. 2015. Di- versifying restricted Boltzmann machine for docu- ment modeling. In KDD.",
                "links": null
            },
            "BIBREF50": {
                "ref_id": "b50",
                "title": "Distilled wasserstein learning for word embedding and topic modeling",
                "authors": [
                    {
                        "first": "Hongteng",
                        "middle": [],
                        "last": "Xu",
                        "suffix": ""
                    },
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Wei",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "NIPS",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hongteng Xu, Wenlin Wang, Wei Liu, and Lawrence Carin. 2018. Distilled wasserstein learning for word embedding and topic modeling. In NIPS.",
                "links": null
            },
            "BIBREF51": {
                "ref_id": "b51",
                "title": "Improved variational autoencoders for text modeling using dilated convolutions",
                "authors": [
                    {
                        "first": "Zichao",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    },
                    {
                        "first": "Zhiting",
                        "middle": [],
                        "last": "Hu",
                        "suffix": ""
                    },
                    {
                        "first": "Ruslan",
                        "middle": [],
                        "last": "Salakhutdinov",
                        "suffix": ""
                    },
                    {
                        "first": "Taylor",
                        "middle": [],
                        "last": "Berg-Kirkpatrick",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1702.08139"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Zichao Yang, Zhiting Hu, Ruslan Salakhutdinov, and Taylor Berg-Kirkpatrick. 2017. Improved varia- tional autoencoders for text modeling using dilated convolutions. arXiv preprint arXiv:1702.08139.",
                "links": null
            },
            "BIBREF52": {
                "ref_id": "b52",
                "title": "SeqGAN: Sequence generative adversarial nets with policy gradient",
                "authors": [
                    {
                        "first": "Lantao",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    },
                    {
                        "first": "Weinan",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Jun",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Yong",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "AAAI",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lantao Yu, Weinan Zhang, Jun Wang, and Yong Yu. 2017. SeqGAN: Sequence generative adversarial nets with policy gradient. In AAAI.",
                "links": null
            },
            "BIBREF53": {
                "ref_id": "b53",
                "title": "Variational neural machine translation",
                "authors": [
                    {
                        "first": "Biao",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Deyi",
                        "middle": [],
                        "last": "Xiong",
                        "suffix": ""
                    },
                    {
                        "first": "Jinsong",
                        "middle": [],
                        "last": "Su",
                        "suffix": ""
                    },
                    {
                        "first": "Hong",
                        "middle": [],
                        "last": "Duan",
                        "suffix": ""
                    },
                    {
                        "first": "Min",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1605.07869"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Biao Zhang, Deyi Xiong, Jinsong Su, Hong Duan, and Min Zhang. 2016. Variational neural machine trans- lation. arXiv preprint arXiv:1605.07869.",
                "links": null
            },
            "BIBREF54": {
                "ref_id": "b54",
                "title": "Sequence generation with guider network",
                "authors": [
                    {
                        "first": "Ruiyi",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Changyou",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Zhe",
                        "middle": [],
                        "last": "Gan",
                        "suffix": ""
                    },
                    {
                        "first": "Wenlin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Liqun",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Guoyin",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1811.00696"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Ruiyi Zhang, Changyou Chen, Zhe Gan, Wenlin Wang, Liqun Chen, Dinghan Shen, Guoyin Wang, and Lawrence Carin. 2018. Sequence generation with guider network. arXiv preprint arXiv:1811.00696.",
                "links": null
            },
            "BIBREF55": {
                "ref_id": "b55",
                "title": "Learning structural weight uncertainty for sequential decision-making",
                "authors": [
                    {
                        "first": "Ruiyi",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Chunyuan",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Changyou",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1801.00085"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Ruiyi Zhang, Chunyuan Li, Changyou Chen, and Lawrence Carin. 2017a. Learning structural weight uncertainty for sequential decision-making. arXiv preprint arXiv:1801.00085.",
                "links": null
            },
            "BIBREF56": {
                "ref_id": "b56",
                "title": "Adversarial feature matching for text generation",
                "authors": [
                    {
                        "first": "Yizhe",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Zhe",
                        "middle": [],
                        "last": "Gan",
                        "suffix": ""
                    },
                    {
                        "first": "Kai",
                        "middle": [],
                        "last": "Fan",
                        "suffix": ""
                    },
                    {
                        "first": "Zhi",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Ricardo",
                        "middle": [],
                        "last": "Henao",
                        "suffix": ""
                    },
                    {
                        "first": "Dinghan",
                        "middle": [],
                        "last": "Shen",
                        "suffix": ""
                    },
                    {
                        "first": "Lawrence",
                        "middle": [],
                        "last": "Carin",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1706.03850"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yizhe Zhang, Zhe Gan, Kai Fan, Zhi Chen, Ricardo Henao, Dinghan Shen, and Lawrence Carin. 2017b. Adversarial feature matching for text generation. arXiv preprint arXiv:1706.03850.",
                "links": null
            },
            "BIBREF57": {
                "ref_id": "b57",
                "title": "Learning discourse-level diversity for neural dialog models using conditional variational autoencoders",
                "authors": [
                    {
                        "first": "Tiancheng",
                        "middle": [],
                        "last": "Zhao",
                        "suffix": ""
                    },
                    {
                        "first": "Ran",
                        "middle": [],
                        "last": "Zhao",
                        "suffix": ""
                    },
                    {
                        "first": "Maxine",
                        "middle": [],
                        "last": "Eskenazi",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1703.10960"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Tiancheng Zhao, Ran Zhao, and Maxine Eskenazi. 2017. Learning discourse-level diversity for neural dialog models using conditional variational autoen- coders. arXiv preprint arXiv:1703.10960.",
                "links": null
            },
            "BIBREF58": {
                "ref_id": "b58",
                "title": "Selective encoding for abstractive sentence summarization",
                "authors": [
                    {
                        "first": "Qingyu",
                        "middle": [],
                        "last": "Zhou",
                        "suffix": ""
                    },
                    {
                        "first": "Nan",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    },
                    {
                        "first": "Furu",
                        "middle": [],
                        "last": "Wei",
                        "suffix": ""
                    },
                    {
                        "first": "Ming",
                        "middle": [],
                        "last": "Zhou",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1704.07073"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Qingyu Zhou, Nan Yang, Furu Wei, and Ming Zhou. 2017. Selective encoding for abstractive sentence summarization. arXiv preprint arXiv:1704.07073.",
                "links": null
            },
            "BIBREF59": {
                "ref_id": "b59",
                "title": "Texygen: A benchmarking platform for text generation models",
                "authors": [
                    {
                        "first": "Yaoming",
                        "middle": [],
                        "last": "Zhu",
                        "suffix": ""
                    },
                    {
                        "first": "Sidi",
                        "middle": [],
                        "last": "Lu",
                        "suffix": ""
                    },
                    {
                        "first": "Lei",
                        "middle": [],
                        "last": "Zheng",
                        "suffix": ""
                    },
                    {
                        "first": "Jiaxian",
                        "middle": [],
                        "last": "Guo",
                        "suffix": ""
                    },
                    {
                        "first": "Weinan",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Jun",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Yong",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1802.01886"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yaoming Zhu, Sidi Lu, Lei Zheng, Jiaxian Guo, Weinan Zhang, Jun Wang, and Yong Yu. 2018. Texy- gen: A benchmarking platform for text generation models. arXiv preprint arXiv:1802.01886.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "num": null,
                "text": "7 b d 2 7 N W 5 6 p q o w 6 H c A Q n 4 M I 5 d O A G u u A B g R y e 4 R X e r C f r x X q 3 P u a j N a v a O Y A / s D 5 / A D S J k v 0 = < / l a t e x i t > t e x i t s h a 1 _ b a s e 6 4 = \" W ST s Q o O p j L 3 b S L 2 I G 0 A M B S D f k 7 I = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 9 L B b B U 0 l E U G 9 F L x 4 r G Ft o Q 9 l s N + 3 S z S b s T s Q S + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m C T T j P s s k Y l u h 9 R w K R T 3 U a D k 7",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF1": {
                "num": null,
                "text": "N a 4 L t I o w x E c w y l 4 c A E N u I U m + M B A w D O 8 w p u j n B f n 3 f m Y t 5 a c Y u Y Q / s D 5 / A E 7 n I 5 / < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W S T s Q o O p j L 3 b S L 2 I G 0 A M B S D f k 7 I = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T s Q S + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m C T T j P s s k Y l u h 9 R w K R T 3 U a D k 7",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF2": {
                "num": null,
                "text": "N a 4 L t I o w x E c w y l 4 c A E N u I U m + M B A w D O 8 w p u j n B f n 3 f m Y t 5 a c Y u Y Q / s D 5 / A E 7 n I 5 / < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W S T s Q o O p j L 3 b S L 2 I G 0 A M B S D f k 7 I = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T s Q S + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m C T T j P s s k Y l u h 9 R w K R T 3 U a D k 7",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF3": {
                "num": null,
                "text": "N a 4 L t I o w x E c w y l 4 c A E N u I U m + M B A w D O 8 w p u j n B f n 3 f m Y t 5 a c Y u Y Q / s D 5 / A E 7 n I 5 / < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" W S T s Q o O p j L 3 b S L 2 I G 0 A M B S D f k 7 I = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 U r 1 q / q h 6 9 L B b B U 0 l E U G 9 F L x 4 r G F t o Q 9 l s N + 3 S z S b s T s Q S + h u 8 e F D x 6 h / y 5 r 9 x 2 + a g r Q 8 G H u / N M D M v T K U w 6 L r f T m l l d W 1 9 o 7 x Z 2 d r e 2 d 2 r 7 h 8 8 m C T T j P s s k Y l u h 9 R w K R T 3 U a D k 7",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF4": {
                "num": null,
                "text": "N a 4 L t I o w x E c w y l 4 c A E N u I U m + M B A w D O 8 w p u j n B f n 3 f m Y t 5 a c Y u Y Q / s D 5 / A E 7 n I 5 / < / l a t e x i t >x 1 < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 2 5 R G + R 1 T c Y q c S h T + o n W Y Y Q e J c E = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x W N F Y w t t K J v t p F 2 6 2 Y T d j V h C f 4 I X D y p e / U f e / D d u 2 x y 0 9 c H A 4 7 0 Z Z u a F q e D a u O 6 3 s 7 S 8 s r q 2 X t o o b 2 5 t 7 + x W 9v Y f d J I p h j 5 L R K J a I d U o u E T f c C O w l S q k c S i w G Q 6 v J 3 7 z E Z X m i b w 3 o x S D m P Y l j z i j x k p 3 T 1 2 v W 6 m 6 N X c K s k i 8 g l S h Q K N b + e r 0 E p b F K A 0 T V O u 2 5 6 Y m y K k y n A k c l z u Z x p S y I e 1 j 2 1 J J Y 9 R B P j 1 1 T I 6 t 0 i N R o m x J Q 6 b q 7 4 m c x l q P 4 t B 2 x t Q M 9 L w 3 E f / z 2 p m J L o K c y z Q z K N l s U Z Q J Y h I y + Z v 0 u E J m x M g S y h S 3 t x I 2 o I o y Y 9 M p 2 x C 8 + Z c X i X 9 a u 6 x 5 t 2 f V + l W R R g k O 4 Q h O w I N z q M M N N M A H B n 1 4 h l d 4 c 4T z 4 r w 7 H 7 P W J a e Y O Y A / c D 5 / A H l 1 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 25 R G + R 1 T c Y q c S h T + o n W Y Y Q e J c E = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x W N F Y w t t K J v t p F 2 6 2 Y T d j V h C f 4 I X D y p e / U f e / D d u 2 x y 0 9 c H A 4 7 0 Z Z u a F q e D a u O 6 3 s 7 S 8 s r q 2 X t o o b 2 5 t 7 + x W 9 v Y f d J I p h j 5 L R K J a I d U o u E T f c C O w l S q k c S i w G Q 6 v J 3 7 z E Z X m i b w 3 o x S D m P Y l j z i j x k p 3 T 1 2 v W 6 m 6 N X c K s k i 8 g l S h Q K N b + e r 0 E p b F K A 0 T V O u 2 5 6 Y m y K k y n A k c l z u Z x p S y I e 1 j 2 1 J J Y 9 R B P j 1 1 T I 6 t 0 i N R o m x J Q 6 b q 7 4 m c x l q P 4 t B 2 x t Q M 9 L w 3 E f / z 2 p m J L o K c y z Q z K N l s U Z Q J Y h I y + Z v 0 u E J m x M g S y h S 3 t x I 2 o I o y Y 9 M p 2 x C 8 + Z c X i X 9 a u 6 x 5 t 2 f V + l W R R g k O 4 Q h O w I N z q M M N N M A H B n 1 4 h l d 4 c 4T z 4 r w 7 H 7 P W J a e Y O Y A / c D 5 / A H l 1 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 2 5 R G + R 1 T c Y q c S h T + o n W Y Y Q e J c E = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x W N F Y w t t K J v t p F 2 6 2 Y T d j V h C f 4 I X D y p e / U f e / D d u 2 x y 0 9 c H A 4 7 0 Z Z u a F q e D a u O 6 3 s 7 S 8 s r q 2 X t o o b 2 5 t 7 + x W 9 v Y f d J I p h j 5 L R K J a I d U o u E T f c C O w l S q k c S i w G Q 6 v J 3 7 z E Z X m i b w 3 o x S D m P Y l j z i j x k p 3 T 1 2 v W 6 m 6 N X c K s k i 8 g l S h Q K N b + e r 0 E p b F K A 0 T V O u 2 5 6 Y m y K k y n A k c l z u Z x p S y I e 1 j 2 1 J J Y 9 R B P j 1 1 T I 6 t 0 i N R o m x J Q 6 b q 7 4 m c x l q P 4 t B 2 x t Q M 9 L w 3 E f / z 2 p m J L o K c y z Q z K N l s U Z Q J Y h I y + Z v 0 u E J m x M g S y h S 3 t x I 2 o I o y Y 9 M p 2 x C 8 + Z c X i X 9 a u 6 x 5 t 2 f V + l W R R g k O 4 Q h O w I N z q M M N N M A H B n 1 4 h l d 4 c 4 T z 4 r w 7 H 7 P W J a e Y O Y A / c D 5 / A H l 1 j X Q = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y 2 5 R G + R 1 T c Y q c S h T + o n W Y Y Q e J c E = \" > A A A B 6 X i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x W N F Y w t t K J v t p F 2 6 2 Y T d j V h C f 4 I X D y p e / U f e / D d u 2 x y 0 9 c H A 4 7 0 Z Z u a F q e D a u O 6 3 s 7 S 8 s r q 2 X t o o b 2 5 t 7 + x W 9 v Y f d J I p h j 5 L R K J a I d U o u E T f c C O w l S q k c S i w G Q 6 v J 3 7 z E Z X m i b w 3 o x S D m P Y l j z i j x k p 3 T 1 2 v W 6 m 6 N X c K s k i 8 g l S h Q K N b + e r 0 E p b F K A 0 T V O u 2 5 6 Y m y K k y n A k c l z u Z x p S y I e 1 j 2 1 J J Y 9 R B P j 1 1 T I 6 t 0 i N R o m x J Q 6 b q 7 4 m c x l q P 4 t B 2 x t Q M 9 L w 3 E f / z 2 p m J L o K c y z Q z K N l s U Z Q J Y h I y + Z v 0 u E J m x M g S y h S 3 t x I 2 o I o y Y 9 M p 2 x C 8 + Z c X i X 9 a u 6 x 5 t 2 f V + l W R R g k O 4 Q h O w I N z q M M N N M A H B n 1 4 h l d 4 c 4 T z 4 r w 7 H 7 P W J a e Y O Y A / c D 5 / A H l 1 j X Q = < / l a t e x i t >x N < l a t e x i t s h a 1 _ b a s e 6 4 = \" W x + R J U e i k j N u 4 4 x i q x C 2 D v c D j 8 o = \" > A A A B 6 3 i c b V B N S 8 N A E J 3 4 W e t X 1 a O X x S J 4 K o k I 6 q 3 o x Z N U M L b Q h r L Z b t q l m 0 3 Y n Y g l 9 D d 4 8 a D i 1 T / k z X / j t s 1 B W x 8 M P N 6 b Y W Z e m E p h 0 H W / n a X l l d W 1 9 d J G e X N r e 2 e 3 s r f / Y J J M M + 6 z R C a 6 F V L D p V D c R 4 G S t 1 L N a R x K 3 g y H 1 x O / + c i 1 E Y m 6 x 1 H K g 5 j 2 l Y g E o 2 g l / 6 m b 3 4 6 7 l a p b c 6 c g",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF5": {
                "num": null,
                "text": "4 h l d 4 c 6 T z 4 r w 7 H / P W F a e Y O Y I / c D 5 / A H x 9 j X Y = < / l a t e x i t > y M < l a t e x i t s h a 1 _ b a s e 6 4 = \" R 2 Y A",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF6": {
                "num": null,
                "text": "m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6 / M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y Va 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF7": {
                "num": null,
                "text": "m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6 / M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y Va 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF8": {
                "num": null,
                "text": "m i n B m f / y I n F P W h c t 5 / a 0 2 b 4 s 2 6 i C f X A A j o A D z k A b X I M O c A E G G X g G r + D N e r J e r H f r Y z Z a s c q d B v g D 6 / M H z i 2 U a g = = < / l a t e x i t > < l a t e x i t s h a 1 _ b a s e 6 4 = \" Y /C 3 2 p N a h B w a h O b g c W V n D h / V M R Y = \" > A A A B + 3 i c b V B P S 8 M w H E 3 n v z n / T X f 0 E h y C F 0 c r g n o b e h G 8 T L B u s J W S p u k W l i Y l S Y Va 5 l f x 4 k H F q 1 / E m 9 / G d O t B N x + E P N 7 7 / c j L C x J G l b b t b 6 u y t L y y u l Z d r 2 1 s b m 3 v 1 H f 3 7 p",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF9": {
                "num": null,
                "text": "D r K e X b e n P d F a 8 U p Z w 7 h F 5 y P b z 6 n j o E = < / l a t e x i t >Attention(b) The extension to text summarization.",
                "uris": null,
                "type_str": "figure"
            },
            "FIGREF10": {
                "num": null,
                "text": "Illustration of the proposed Topic-Guided Variational Autoencoder (TGVAE) for text generation. (a) For generation (the red arrows), the topics inferred from a neural topic model are used to guide a Gaussian mixture prior of the latent code, which is further fed into the decoder to generate a sentence. For inference (the black arrows), the sentence is encoded into a vector and then propagated through the Householder flow to obtain the approximate posterior. (b) An attention module is further added for text summarization.",
                "uris": null,
                "type_str": "figure"
            },
            "TABREF0": {
                "num": null,
                "html": null,
                "text": "ABS 29.55 11.32 26.42 26.55 7.06 22.05 ABS+ 29.78 11.89 26.97 28.18 8.49 23.81 RAS-LSTM 32.55 14.70 30.03 28.97 8.26 24.06 RAS-Elman 33.78 15.97 31.15 27.41 7.69 23.06 lvt2k-lsent 32.67 15.59 30.64 28.35 9.46 24.59 lvt5k-lsent 35.30 16.64 32.62 28.61 9.42 25.24 ASC+FSC 34.17 15.94 31.92 26.73 8.39 23.88 Seq2Seq 34.03 15.93 31.68 28.39 9.26 24.83 Var-Seq2Seq 34.00 15.97 31.85 28.11 9.24 24.86 Var-Seq2Seq-HF (K=1) 34.04 15.98 31.84 28.18 9.27 24.84 Var-Seq2Seq-HF (K=10) 34.22 16.10 32.13 28.78 9.11 24.96 TGVAE (K=0, T=10) 35.34 16.68 32.69 28.99 9.21 24.89 TGVAE (K=1, T=10) 35.35 16.70 32.64 29.02 9.24 24.93 TGVAE (K=10, T=10) 35.40 16.77 32.71 29.07 9.32 25.17 TGVAE (K=10, T=30) 35.59 17.18 32.89 29.38 9.60 25.22 TGVAE (K=10, T=50) 35.63 17.27 33.02 29.65 9.55 25.38",
                "content": "<table><tr><td>DUC-2004</td></tr><tr><td>RF-1 RF-2 RF-L RR-1 RR-2 RR-L</td></tr></table>",
                "type_str": "table"
            },
            "TABREF1": {
                "num": null,
                "html": null,
                "text": "Results on Gigawords and DUC-2004.",
                "content": "<table/>",
                "type_str": "table"
            },
            "TABREF4": {
                "num": null,
                "html": null,
                "text": "10 topics learned from our model on APNEWS, IMDB, BNC and Gigawords.",
                "content": "<table/>",
                "type_str": "table"
            }
        }
    }
}