TMLR-Group-HF

/

Majority-Voting-Qwen3-8B-Base-DAPO14k

like 2

Follow

TMLR Group 17

This PR significantly enhances the model card for Majority-Voting: Qwen3-8B-Base trained on DAPO-14k by adding crucial metadata and expanding its content.

The changes include:

• Adding pipeline_tag: text-generation to reflect the model's function as a Large Language Model for reasoning tasks, improving its discoverability on the Hub.
• Adding library_name: transformers as the model is based on the Qwen3ForCausalLM architecture and is compatible with the Hugging Face Transformers library. This enables the automated "Use in Transformers" widget.
• Adding tags: [reasoning] to highlight the model's core application area, as detailed in the paper.
• Updating the model card's main content to explicitly link to the paper: Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models.
• Including the paper's abstract to provide users with a comprehensive overview of the research and methodology.
• Integrating the detailed "Checkpoints" tables directly from the GitHub repository, which list all related models, their sizes, and methods, offering valuable context and facilitating further exploration.
• Adding the academic citation for proper attribution.

Per the instructions, a sample usage section was not included as no direct, straightforward inference code snippet for this specific model using transformers was found in the GitHub README.

Please review these improvements.