Review Model Card

A model card is structured documentation describing a model's intended uses, evaluation results, limitations, relevant training or development context, and considerations for responsible use.

It is a reference sheet for understanding what a model was built and tested to do, where evidence is weak, and what users or deployers should consider.

Model cards make important claims inspectable, but their usefulness depends on completeness, current versions, honest limitations, and connection to the deployed system.

A model card lists supported languages, evaluation groups, known failure patterns, prohibited uses, version information, and contact details for reporting issues.

A model card is not a certification or guarantee, and documentation about a base model does not describe every application built around it.

A model card states what was tested, intended, and still uncertain.

* [[capability-provenance|Capability Provenance]] * [[data-provenance|Data Provenance]] * [[model-risk-management|Model Risk Management]] * [[due-diligence|Due Diligence]] * [[interpretability-explainability|Interpretability and Explainability]]

* [Mitchell et al., Model Cards for Model Reporting](https://research.google/pubs/model-cards-for-model-reporting/) — primary paper proposing structured model documentation.

* Use model documentation as evidence within [[due-diligence|Due Diligence]], not as a substitute for it.