---
base_model:
- Qwen/Qwen2.5-7B-Instruct
language:
- en
license: apache-2.0
metrics:
- trec_eval
library_name: transformers
pipeline_tag: text-ranking
---

This is a reasoning reranking agent model built upon Qwen-2.5-7B for the paper [REARANK: Reasoning Re-ranking Agent via Reinforcement Learning](https://huggingface.co/papers/2505.20046). The model is trained on [reranking dataset](https://huggingface.co/datasets/le723z/rearank_12k) built from only 179 queries using GRPO to perform reranking task, the codebase is at https://github.com/lezhang7/Rearank

![image/png](https://cdn-uploads.huggingface.co/production/uploads/633b423e5df91da9ceafe40a/-gxwZPWs-AdxunM5wl7Wg.png)

![image/png](https://cdn-uploads.huggingface.co/production/uploads/633b423e5df91da9ceafe40a/PvYue0TOB8ClcV0u0oxeS.png)