jasonrqh

/

InternLM2.5-20B_Math-CoT-20k_lr5e-5_ep8_bs256

like 0

This PR adds a model card for the research presented in "Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability".

The model card includes:

• A link to the paper on Hugging Face Papers.
• A link to the official GitHub repository.
• A summary of the core findings, including the dip-and-recovery pattern and the role of base model capability.
• Metadata for the text-generation pipeline and transformers library.
• The official BibTeX citation.