Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding?

Robust-U1 is a unified Multimodal Large Language Model (MLLM) that can self-recover corrupted visual content and perform multimodal reasoning over it. This enables robust visual understanding even under real-world image degradations.

• Paper: Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding?
• Repository: jqtangust/Robust-U1
• Demo: Hugging Face Space

🏰 Pretrained Checkpoints

Checkpoint	Link	Note
BAGEL-7B-MoT	ByteDance-Seed/BAGEL-7B-MoT	Used as initial weights for training.
Robust-U1	Jiaqi-hkust/Robust-U1	Final model for visual self-recovery and multimodal reasoning.
Robust-U1-RL	Jiaqi-hkust/Robust-U1-RL	Fine-tuned with reinforcement learning.
Robust-U1-SFT	Jiaqi-hkust/Robust-U1-SFT	Fine-tuned with supervised learning.

⭐️ Citation

If you find this repository useful, please cite the paper:

@inproceedings{tang2026robustu1,
      title={Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding?},
      author={Tang, Jiaqi and Chen, Jianmin and Zhai, Youyang and Wei, Wei and Liu, Runtao and Zhao, Mengjie and Wu, Xiangyu and Xiao, Qingfa and Chen, Qifeng},
      booktitle={Proceedings of the 43rd International Conference on Machine Learning (ICML)},
      year={2026},
}