RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback Paper • 2312.00849 • Published Dec 1, 2023 • 8