Jiaqi-hkust/Robust-R1-RL

Robust-R1:Degradation-Aware Reasoning for Robust Visual Understanding

🏰 Pretrained and Fine-tuned Model

• The following checkpoints are utilized to run Robust-R1:

  Checkpoint	Link	Note
  Qwen2.5-VL-Base	link	Used as initial weights for training.
  Robust-R1-SFT	link	Fine-tuned on Robust-R1 dataset
  Robust-R1-RL	link	Fine-tuned with reinforcement learning on Robust-R1 dataset

⭐️ Citation

If you find Robust-R1 useful for your research and applications, please cite using this BibTeX:

@inproceedings{tang2025robustr1,
  title={Robust-R1: Degradation-Aware Reasoning for Robust Visual Understanding},
  author={Tang, Jiaqi and Chen, Jianmin and Wei, Wei and Xu, Xiaogang and Liu, Runtao and Wu, Xiangyu and Xie, Qipeng and Wu, Jiafei and Zhang, Lei and Chen, Qifeng},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2026}
}