RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback Project page for RLHF-V The code is borrowed from VPGTrans and MiniGPT-4. Thanks for the excellent code!