VL-Rethinker-72B

VL-Rethinker-72B achieves SoTA results on various multimodal reasoning benchmarks.

It is trained using the Forced Rethinking technique, on top of VL-Reasoner with GRPO-SSR training.

For details of our approach and performance comparison, please see our paper.

For details of training and evaluation, please see our code repo.

Explore further via the following links:

| 🚀Project Page | 📖Paper | 🔗Github | 🤗Data (Coming Soon) |

Citation

If you feel this model useful, please give us a free cite:

@article{vl-rethinker,
      title={VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning},
      author = {Wang, Haozhe and Qu, Chao and Huang, Zuming and Chu, Wei and Lin, Fangzhen and Chen, Wenhu},
      journal={arXiv preprint arXiv:2504.08837},
      year={2025}
}