arxiv:2505.20046

REARANK: Reasoning Re-ranking Agent via Reinforcement Learning

Published on May 26

· Submitted by

le723z on May 27

Upvote

Authors:

Le Zhang ,

Abstract

REARANK, a reinforcement learning-enhanced large language model for listwise reasoning,outperforms baseline models and even surpasses GPT-4 on reasoning-intensive benchmarks with minimal data.

AI-generated summary

We present REARANK, a large language model (LLM)-based listwise reasoning reranking agent. REARANK explicitly reasons before reranking, significantly improving both performance and interpretability. Leveraging reinforcement learning and data augmentation, REARANK achieves substantial improvements over baseline models across popular information retrieval benchmarks, notably requiring only 179 annotated samples. Built on top of Qwen2.5-7B, our REARANK-7B demonstrates performance comparable to GPT-4 on both in-domain and out-of-domain benchmarks and even surpasses GPT-4 on reasoning-intensive BRIGHT benchmarks. These results underscore the effectiveness of our approach and highlight how reinforcement learning can enhance LLM reasoning capabilities in reranking.

View arXiv page View PDF GitHub repository Add to collection

Community

le723z

Paper author Paper submitter 10 days ago

We present Rearank, a large language model (LLM)-based listwise reasoning reranking agent. Rearank explicitly reasons before reranking, significantly improving both performance and interpretability. Leveraging reinforcement learning and data augmentation, Rearank achieves substantial improvements over baseline models across popular information retrieval benchmarks, notably requiring only 179 annotated samples. Built on top of Qwen2.5-7B, our Rearank-7B demonstrates performance comparable to GPT-4 on both in-domain and out-of-domain benchmarks and even surpasses GPT-4 on reasoning-intensive BRIGHT benchmarks. These results underscore the effectiveness of our approach and highlight how reinforcement learning can enhance LLM reasoning capabilities in reranking. The code is available https://github.com/lezhang7/Rearank.