hugging-face-evaluation

To convert README evaluation tables to Hugging Face model-index format, inspect the model repository's README to identify tables, extract the desired table to YAML using the task type, and apply the changes to generate validated model-index entries for the model card widget.

Yes, you can import Artificial Analysis benchmark scores into a Hugging Face model card by providing your AA_API_KEY and creator slug, fetching the evaluations, and using the import command to merge them with existing model-index data via a pull request.

Yes, the skill supports running custom model evaluations using vLLM or lighteval for GPU-backed workflows, allowing you to generate new benchmark scores and automatically push the updated evaluation results to model cards via pull requests or direct commits.

The best way to automate pull-request updates for Hugging Face model cards is to use a script that extracts README evaluations or imports Artificial Analysis data, converts it to model-index format, and submits the validated changes directly to the repository via a PR.

Yes, you need an AA_API_KEY to import Artificial Analysis evaluations into model cards, which authorizes the script to fetch benchmark scores for your specific creator slug and model name before merging them into the Hugging Face model-index format.

Validating the model-index format ensures that README-stored scores and API-sourced benchmarks are correctly transformed into standardized entries, preventing format errors and enabling seamless, automated updates to Hugging Face model cards without manual correction.