how2everything

/

how2judge

like 0

This PR adds a model card for How2Judge, the 8B judge model described in the paper How2Everything: Mining the Web for How-To Procedures to Evaluate and Improve LLMs.

The model card includes:

• Metadata for pipeline_tag, library_name, and license.
• A description of the model's purpose and its performance (80.5% agreement with human annotators).
• Links to the paper and the official GitHub repository.
• The BibTeX citation from the paper.