ulab-ai

/

sotopia-rl-qwen2.5-7B-rm

like 1

Follow

ulab 30

This PR significantly enhances the model card for ulab-ai/sotopia-rl-qwen-2.5-7B-grpo by:

• Adding the pipeline_tag: feature-extraction, which is appropriate for a reward model that outputs scores/features from text.
• Specifying library_name: transformers as the primary library for model interaction, while also adding peft to the general tags to indicate its adapter nature.
• Including relevant tags such as reward-model, social-intelligence, and reinforcement-learning for better discoverability.
• Linking directly to the associated paper: Sotopia-RL: Reward Design for Social Intelligence.
• Adding links to the official project page: https://rl.sotopia.world.
• Providing a direct link to the GitHub repository: https://github.com/sotopia-lab/sotopia-rl.
• Expanding the model description with an abstract and introduction.
• Including a practical Python usage example for loading the model and performing inference as a sequence classification model.
• Adding a citation section for the paper.

These improvements will make the model more informative and user-friendly on the Hugging Face Hub.