File size: 1,510 Bytes
e8f7c7f b1e50f1 e8f7c7f b1e50f1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 |
---
license: apache-2.0
tags:
- Reinforcement Learning
- Visual-langauge Reasoning
---
# Model Card for WeThink-Qwen2.5VL-7B
Repository: https://github.com/yangjie-cv/WeThink
Paper: https://arxiv.org/abs/2506.07905
## 🏆 Performance Highlights
**WeThink-Qwen2.5VL-7B** achieves:
- 🥇 **1st place** on [OpenCompass Multimodal Reasoning Leaderboard](https://rank.opencompass.org.cn/leaderboard-multimodal-reasoning/?m=REALTIME)
- 🏅 **5th place** on [OpenCompass Multi-modal Academic Leaderboard](https://rank.opencompass.org.cn/leaderboard-multimodal/?m=REALTIME)
*(As of May 30th, 2025)*
## 🚀 Quick Start
### Inference
```bash
git clone https://github.com/yangjie-cv/WeThink
cd WeThink
python inference.py
```
💡 Note: System prompt is required during inference.
### 📊 Evaluation
We have integrated WeThink-Qwen2.5VL-7B into the [VLMEvalKit](https://github.com/open-compass/VLMEvalKit). Please follow its [Quickstart guide](https://github.com/open-compass/VLMEvalKit/blob/main/docs/en/Quickstart.md) to evaluate WeThink-Qwen2.5VL-7B on various benchmarks.
## Citation
```
@misc{yang2025wethink,
title={WeThink: Toward General-purpose Vision-Language Reasoning via Reinforcement Learning},
author={Jie Yang and Feipeng Ma and Zitian Wang and Dacheng Yin and Kang Rong and Fengyun Rao and Ruimao Zhang},
year={2025},
eprint={2506.07905},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2506.07905},
}
``` |