jinzhuoran

/

OmniRewardModel

like 5

This PR enhances the model card for Omni-Reward by:

• Adding library_name: transformers to correctly enable the "how to use" widget, as the model's file structure (e.g., config.json, tokenizer_config.json) indicates compatibility with the Hugging Face Transformers library.
• Setting pipeline_tag: any-to-any to accurately reflect its omni-modal capabilities (handling text, image, video, audio, and 3D data) and improve discoverability on the Hub.
• Including a direct link to the paper (Omni-Reward: Towards Generalist Omni-Modal Reward Modeling with Free-Form Preferences) and the official GitHub repository (https://github.com/HongbangYuan/OmniReward) in the introductory link section.
• Correcting a minor </a></a> typo in the existing benchmark link.

These updates improve the model's discoverability and provide users with comprehensive information and usage guidance.