TMLR-Group-HF

/

Majority-Voting-Llama-3.2-3B-Instruct-DAPO14k

like 0

Follow

TMLR Group 20

This PR significantly enhances the model card for the Majority-Voting: Llama-3.2-3B-Instruct model.

The key improvements include:

• Metadata: Adding pipeline_tag: text-generation for better discoverability, library_name: transformers to enable the automated "how to use" widget on the Hub (supported by config.json and internal usage in the provided snippet), and relevant tags such as llama, fine-tuned, and mathematical-reasoning.
• Paper Link: A direct link to the paper "Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models" is added.
• Expanded Description: A detailed overview of the model and the "Co-rewarding" framework, derived from the paper's abstract, is now included.
• GitHub Repository Link: The existing GitHub link has been updated and made more prominent.
• Sample Usage: A "How to use" section with a Python code snippet, directly sourced from the project's GitHub README, is added to provide immediate guidance on model inference. This snippet includes the literal \n characters as found in the original source, as per instructions.

These changes collectively make the model card more informative and user-friendly.