Code for our paper mDPO: Conditional Preference Optimization for Multimodal Large Language Models. Please check our website for an overview of our work.

• 🔥 [2024-09-04] Initial release of the mDPO trainer. We are currently working on releasing the code for training and evaluating different models.

TBD

To train Bunny with mDPO, use the following command:

python bunny/run_mdpo_bunny.py

TBD

Please cite the following paper if you find the repo helpful:

@article{wang2024mdpo,
  title={mDPO: Conditional Preference Optimization for Multimodal Large Language Models},
  author={Wang, Fei and Zhou, Wenxuan and Huang, James Y and Xu, Nan and Zhang, Sheng and Poon, Hoifung and Chen, Muhao},
  journal={arXiv preprint arXiv:2406.11839},
  year={2024}
}