SFEvent (Open-Source AI Meetup)

posted an update 13 days ago

Post

946

🚀 Just Dropped: MathX-5M — Your Gateway to Math-Savvy GPTs

👨‍🔬 Wanna fine-tune your own GPT for math?
🧠 Building a reasoning agent that actually *thinks*?
📊 Benchmarking multi-step logic across domains?

Say hello to [**MathX-5M**]( XenArcAI/MathX-5M) — a **5 million+ sample** dataset crafted for training and evaluating math reasoning models at scale.

Built by **XenArcAI**, it’s optimized for:
- 🔍 Step-by-step reasoning with , , and formats
- 🧮 Coverage from arithmetic to advanced algebra and geometry
- 🧰 Plug-and-play with Gemma, Qwen, Mistral, and other open LLMs
- 🧵 Compatible with Harmony, Alpaca, and OpenChat-style instruction formats

Whether you're prototyping a math tutor, testing agentic workflows, or just want your GPT to solve equations like a pro—**MathX-5M is your launchpad**.

🔗 Dive in: ( XenArcAI/MathX-5M)

Let’s make open-source models *actually* smart at math.
#FineTuneYourGPT #MathX5M #OpenSourceAI #LLM #XenArcAI #Reasoning #Gemma #Qwen #Mistral

merterbak

posted an update 14 days ago

Post

2433

OpenAI is now open again! Check out OpenAI’s brand new gpt‑oss‑20b model hosted on ZeroGPU 🤗

merterbak/gpt-oss-20b-demo

Parveshiiii

posted an update 20 days ago

Post

1037

🚀 Launch Alert: Dev-Stack-Agents
Meet your 50-agent senior AI team — principal-level experts in engineering, AI, DevOps, security, product, and more — all bundled into one modular repo.

+ Code. Optimize. Scale. Secure.
- Full-stack execution, Claude-powered. No human bottlenecks.

🔧 Built for Claude Code
Seamlessly plug into Claude’s dev environment:

* 🧠 Each .md file = a fully defined expert persona
* ⚙️ Claude indexes them as agents with roles, skills & strategy
* 🤖 You chat → Claude auto-routes to the right agent(s)
* ✍️ Want precision? Just call @agent-name directly
* 👥 Complex task? Mention multiple agents for team execution

Examples:

"@security-auditor please review auth flow for risks"
"@cloud-architect + @devops-troubleshooter → design a resilient multi-region setup"
"@ai-engineer + @legal-advisor → build a privacy-safe RAG pipeline"

🔗 https://github.com/Parveshiiii/Dev-Stack-Agents
MIT License | Claude-Ready | PRs Welcome

1 reply

·

Parveshiiii

posted an update about 1 month ago

Post

2676

🧠 Glimpses of AGI — A Vision for All Humanity
What if AGI wasn’t just a distant dream—but a blueprint already unfolding?

I’ve just published a deep dive called Glimpses of AGI, exploring how scalable intelligence, synthetic reasoning, and alignment strategies are paving a new path forward. This isn’t your average tech commentary—it’s a bold vision for conscious AI systems that reason, align, and adapt beyond narrow tasks.

🔍 Read it, upvote it if it sparks something, and let’s ignite a collective conversation about the future of AGI.

https://huggingface.co/blog/Parveshiiii/glimpses-of-agi

Parveshiiii

posted an update about 2 months ago

Post

2826

🧠 MathX-5M by XenArcAI — Scalable Math Reasoning for Smarter LLMs

Introducing MathX-5M, a high-quality, instruction-tuned dataset built to supercharge mathematical reasoning in large language models. With 5 million rigorously filtered examples, it spans everything from basic arithmetic to advanced calculus—curated from public sources and enhanced with synthetic data.

🔍 Key Highlights:
- Step-by-step reasoning with verified answers
- Covers algebra, geometry, calculus, logic, and more
- RL-validated correctness and multi-stage filtering
- Ideal for fine-tuning, benchmarking, and educational AI

📂 - XenArcAI/MathX-5M

1 reply

·

merterbak

posted an update 3 months ago

Post

2773

Qwen 3 technical report released🚀
Report: https://github.com/QwenLM/Qwen3/blob/main/Qwen3_Technical_Report.pdf

merterbak

posted an update 3 months ago

Post

2603

Seed-Coder released and it's designed for coding tasks, featuring base, instruct, and reasoning variants at an 8B parameter scale developed by ByteDance Seed team. Unlike traditional open source LLMs that rely on human crafted rules or annotated data for curating code pretraining datasets Seed-Coder introduces a model-centric data pipeline. The pipeline processes raw data from GitHub and web archives into four categories: file-level codes, repository-level codes, GitHub commits, and code-related web data.A quality filter LLM, evaluates code (for readability, modularity, clarity, and reusability) by removing the lowest 10% to create a 6 trillion token dataset supporting 89 programming languages.
Models: ByteDance-Seed/seed-coder-680de32c15ead6555c75b0e4
Github: https://github.com/ByteDance-Seed/Seed-Coder/tree/master
Paper: https://github.com/ByteDance-Seed/Seed-Coder/blob/master/Seed-Coder.pdf

merterbak

posted an update 4 months ago

Post

1707

Microsoft released their new fine-tuned phi-4 models with reasoning data yesterday. They outperform/rival much larger models . Check out them if you haven't yet. 🚀

Phi4 mini reasoning(SFT): microsoft/Phi-4-mini-reasoning
Phi-4 reasoning(SFT): microsoft/Phi-4-reasoning
Phi-4 reasoning plus (SFT + RL): microsoft/Phi-4-reasoning-plus
Demo: https://github.com/marketplace/models/azureml/Phi-4-reasoning/playground
Articles: https://arxiv.org/pdf/2504.21318
https://arxiv.org/pdf/2504.21233
Blog: https://azure.microsoft.com/en-us/blog/one-year-of-phi-small-language-models-making-big-leaps-in-ai/

1 reply

·

merterbak

posted an update 4 months ago

Post

4911

Qwen 3 models released🔥
It offers 2 MoE and 6 dense models with following parameter sizes: 0.6B, 1.7B, 4B, 8B, 14B, 30B(MoE), 32B, and 235B(MoE).
Models: Qwen/qwen3-67dd247413f0e2e4f653967f
Blog: https://qwenlm.github.io/blog/qwen3/
Demo: Qwen/Qwen3-Demo
GitHub: https://github.com/QwenLM/Qwen3

✅ Pre-trained 119 languages(36 trillion tokens) and dialects with strong translation and instruction following abilities. (Qwen2.5 was pre-trained on 18 trillion tokens.)
✅Qwen3 dense models match the performance of larger Qwen2.5 models. For example, Qwen3-1.7B/4B/8B/14B/32B perform like Qwen2.5-3B/7B/14B/32B/72B.
✅ Three stage done while pretraining:
• Stage 1: General language learning and knowledge building.
• Stage 2: Reasoning boost with STEM, coding, and logic skills.
• Stage 3: Long context training
✅ It supports MCP in the model
✅ Strong agent skills
✅ Supports seamless between thinking mode (for hard tasks like math and coding) and non-thinking mode (for fast chatting) inside chat template.
✅ Better human alignment for creative writing, roleplay, multi-turn conversations, and following detailed instructions.

merterbak

posted an update 4 months ago

Post

3626

FlowReasoner is a new system that builds a custom set of small AI agents for every user question. Unlike search based methods it uses reasoning driven optimization with external execution feedback.

✅ First, it distills reasoning data using DeepSeek R1-671B to build multi agent systems. 🤖
✅ Then, reasoning data used for DeepSeek-R1-Distill-Qwen-7B via supervised fine tuning for basic reasoning skills. 💡
✅ Finally, RL with GRPO (optimizes by comparing response groups from queries/tasks) to improve reasoning.

FlowReasoner: Reinforcing Query-Level Meta-Agents (2504.15257)
Code: https://github.com/sail-sg/flowreasoner

merterbak

posted an update 4 months ago

Post

2236

Here’s a cool paper I found: “Massive Image Embedding Benchmark (MIEB).” It is a new tool to test how good image embedding models are. It has 130 different tasks grouped into 8 categories, like image search, classification, clustering similar images, answering questions based on images, and understanding documents. It even covers 38 different languages.

The authors tested 50 models and found that no single model was best at everything. Some models were great at recognizing text inside images but struggled to handle complicated tasks like matching images and text that appear together.

Paper: https://arxiv.org/pdf/2504.10471v1
Code: https://github.com/embeddings-benchmark/mteb

2 replies

·

merterbak

posted an update 4 months ago

Post

3056

OpenAI published 2 benchmark datasets on Hugging Face 🔥
openai/mrcr
openai/graphwalks
MRCR tests how well a model can find the right answer when many similar questions are spread out in a long context. Graphwalks checks if a model can follow steps in a big graph and find the correct nodes by thinking through the structure

merterbak

posted an update 4 months ago

Post

3433

OpenAI has released BrowseComp an open source benchmark designed to evaluate the web browsing capabilities of AI agents. This dataset comprising 1,266 questions challenges AI models to navigate the web and uncover complex and obscure information. Crafted by human trainers, the questions are intentionally difficult. (unsolvable by another person in under ten minutes and beyond the reach of existing models like ChatGPT with and without browsing and an early version of OpenAI's Deep Research tool.)

Blog Post: https://openai.com/index/browsecomp/
Paper: https://cdn.openai.com/pdf/5e10f4ab-d6f7-442e-9508-59515c65e35d/browsecomp.pdf
Code in simple eval repo: https://github.com/openai/simple-evals

merterbak

posted an update 4 months ago

Post

4766

Qwen 3 can launch very soon. 👀

https://github.com/ggml-org/llama.cpp/pull/12828

3 replies

·

merterbak

posted an update 5 months ago

Post

3041

Meta has unveiled its Llama 4 🦙 family of models, featuring native multimodality and mixture-of-experts architecture. Two model families are available now:
Models🤗: meta-llama/llama-4-67f0c30d9fe03840bc9d0164
Blog Post: https://ai.meta.com/blog/llama-4-multimodal-intelligence/
HF's Blog Post: https://huggingface.co/blog/llama4-release

- 🧠 Native Multimodality - Process text and images in a unified architecture
- 🔍 Mixture-of-Experts - First Llama models using MoE for incredible efficiency
- 📏 Super Long Context - Up to 10M tokens
- 🌐 Multilingual Power - Trained on 200 languages with 10x more multilingual tokens than Llama 3 (including over 100 languages with over 1 billion tokens each)

🔹 Llama 4 Scout
- 17B active parameters (109B total)
- 16 experts architecture
- 10M context window
- Fits on a single H100 GPU
- Beats Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1

🔹 Llama 4 Maverick
- 17B active parameters (400B total)
- 128 experts architecture
- It can fit perfectly on DGX H100(8x H100)
- 1M context window
- Outperforms GPT-4o and Gemini 2.0 Flash
- ELO score of 1417 on LMArena currently second best model on arena

🔹 Llama 4 Behemoth (Coming Soon)
- 288B active parameters (2T total)
- 16 experts architecture
- Teacher model for Scout and Maverick
- Outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks

merterbak

posted an update 6 months ago

Post

3819

🔥 Meet Muse: that can generate a game environment based on visuals or players’ controller actions. It was developed by Microsoft Research in collaboration with Ninja Theory (Hellblade developer). It’s built on something called the World and Human Action Model (WHAM-1.6B model). They trained on 7 years of Bleeding Edge gameplay and it can generate 2 minute long 3D game sequences with consistent physics and character behaviors all from just a second of input. They’ve gone and open-sourced it too. Open weights, the WHAM Demonstrator, and sample data on Azure AI Foundry for anyone to play with. Hope so soon on Hugging Face 🤗.

📄 Paper: https://www.nature.com/articles/s41586-025-08600-3
Blog Post: https://www.microsoft.com/en-us/research/blog/introducing-muse-our-first-generative-ai-model-designed-for-gameplay-ideation/

1 reply

·

Taylor658

posted an update 9 months ago

Post

935

🌐 The Stanford Institute for Human-Centered AI (https://aiindex.stanford.edu/vibrancy/) has released its 2024 Global AI Vibrancy Tool, a way to explore and compare AI progress across 36 countries.

📊 It measures progress across the 8 broad pillars of R&D, Responsible AI, Economy, Education, Diversity, Policy and Governance, Public Opinion and Infrastructure. (Each of these pillars have a number of Sub Indices)

📈 As a whole it is not surprising that the USA was at the top in terms of overall score as of 2023 (AI investment activity is a large part of the economic pillar for example and that is a large part of the overall USA ranking) but drilling in to more STRATEGIC Macro pillars like Education, Infrastructure or R&D reveal interesting growth patterns in Asia (particularly China) and Western Europe that I suspect the 2024 metrics will bear out.

🤖 Hopefully the 2024 Global Vibrancy ranking will break out AI and ML verticals like Computer Vision or NLP and or the AI Agent space as that may also from a global macro level give indications of what is to come globally for AI in 2025.

Taylor658

posted an update 9 months ago

Post

1161

🤖💻 Function Calling is a key component of Agent workflows. To call functions, an LLM needs a way to interact with other systems and run code. This usually means connecting it to a runtime environment that can handle function calls, data, and security.

Per the Berkeley Function-Calling Leaderboard there are only 2 fully open source models (The other 2 in the top 20 that are not closed source have cc-by-nc-4.0 licenses) out of the top 20 models that currently have function calling built in as of 17 Nov 2024.
https://gorilla.cs.berkeley.edu/leaderboard.html

The 2 Open Source Models out of the top 20 that currently support function calling are:

meetkai/functionary-medium-v3.1
Team-ACE/ToolACE-8B

This is a both a huge disadvantage AND an opportunity for the Open Source community as Enterprises, Small Business, Government Agencies etc. quickly adopt Agents and Agent workflows over the next few months. Open Source will have a lot of catching up to do as Enterprises will be hesitant to switch from the closed source models that they may initially build their Agent workflows on in the next few months to an open source alternative later.

Hopefully more open source models will support function calling in the near future.

Draichi

posted an update 10 months ago

Post

3574

🏁 Now it is possible to chat with telemetry data from real Formula 1 races!

This is an AI-powered solution for analyzing and generating detailed reports on Formula 1 racing sessions. This project combines the power of ReAct agents from LangChain with a RAG approach to pull data from a SQL database.

At the core of this system is a text-to-SQL capability that allows users to ask natural language questions about various aspects of F1 races, such as driver performance, weather impact, race strategies, and more. The AI agent then queries the database, processes the information, and generates comprehensive reports tailored to the user's needs.

The reports can be exported in various formats, making it easy to share insights with team members, race fans, or the broader motorsports community.

(The project is in beta, some erros may occur)

Check it out:

- Draichi/Formula1-race-debriefing
- https://github.com/Draichi/formula1-AI

1 reply

·

Taylor658

posted an update 10 months ago

Post

2314

The Mystery Bot 🕵️‍♂️ saga I posted about from earlier this week has been solved...🤗

Cohere for AI has just announced its open source Aya Expanse multilingual model. The Initial release supports 23 languages with more on the way soon.🌌 🌍

You can also try Aya Expanse via SMS on your mobile phone using the global WhatsApp number or one of the initial set of country specific numbers listed below.⬇️

🌍WhatsApp - +14313028498
Germany - (+49) 1771786365
USA – +18332746219
United Kingdom — (+44) 7418373332
Canada – (+1) 2044107115
Netherlands – (+31) 97006520757
Brazil — (+55) 11950110169
Portugal – (+351) 923249773
Italy – (+39) 3399950813
Poland - (+48) 459050281

1 reply

·

Open-Source AI Meetup

AI & ML interests

Recent Activity

AI & ML interests

Recent Activity

Team members 580

SFEvent's activity