https://github.com/RLHF-V/RLHF-V

[CVPR'24] RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback
https://github.com/RLHF-V/RLHF-V

chatbot gpt-4 llama multi-modality multimodal rlhf-v visual-language-learning

Last synced: about 2 months ago
JSON representation

[CVPR'24] RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback

• Host: GitHub
• URL: https://github.com/RLHF-V/RLHF-V
• Owner: RLHF-V
• Created: 2023-11-29T07:32:32.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-09-11T09:21:43.000Z (7 months ago)
• Last Synced: 2024-10-18T21:59:14.525Z (6 months ago)
• Topics: chatbot, gpt-4, llama, multi-modality, multimodal, rlhf-v, visual-language-learning
• Language: Python
• Homepage: https://rlhf-v.github.io
• Size: 70.6 MB
• Stars: 228
• Watchers: 2
• Forks: 6
• Open Issues: 1
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

• awesome-RLHF - official
• StarryDivineSky - RLHF-V/RLHF-V - V项目是CVPR 2024的一项研究，旨在通过细粒度的修正性人类反馈，实现更值得信赖的多模态大型语言模型（MLLMs）。该项目提出了一种行为对齐方法，通过人类提供的细致修正意见来训练模型，使其行为更符合人类期望。核心思想是利用人类反馈来纠正模型在视觉理解和推理方面的错误，从而提高模型的可信度和可靠性。项目关注于提升MLLMs在处理视觉信息时的准确性和一致性，使其能够更好地理解图像并生成更合理的文本描述。这种方法通过对模型行为的微调，使其在复杂场景下能够做出更明智的决策，从而增强用户对模型的信任。简单来说，RLHF-V利用人类的“修改意见”来训练AI，让AI更好地理解图像并做出正确的判断，最终让AI更值得信任。 (多模态大模型 / 资源传输下载)