MODEL-CARD

Model Card AI safety + governance framework.

Définition

Model Card est le AI/ML model documentation standard proposed par Google Research Margaret Mitchell + Timnit Gebru et al. paper 'Model Cards for Model Reporting' 2019, qui standardise documentation des ML models published incluant model details + intended use + factors + metrics + evaluation data + training data + quantitative analyses + ethical considerations + caveats + recommendations, foundational pour responsible AI documentation widely adopted Hugging Face + multiple AI publication platforms. Framework + standard + guidance objectifs AI safety + governance + risk management + transparency + fairness + accountability + explainability + multiple AI ethics principles adoption multi-stakeholder process + voluntary adoption industry + emerging regulation alignment EU AI Act + US Executive Orders + multiple national jurisdictions + private sector adoption Microsoft + Google + Meta + OpenAI + Anthropic + autres major AI labs + corporate compliance teams 2020s+.

Origine

Model Cards paper Mitchell + Gebru et al. published 2019 ; widely adopted Hugging Face + GitHub + Google + Meta + Microsoft AI model publications.

Exemple en contexte

Hugging Face hosts Llama 3 model published Meta : Model Card on Hugging Face includes (1) Model details (Meta + parameters + version + date + license), (2) Intended use cases + limitations, (3) Bias + fairness considerations + evaluation results across demographic groups, (4) Training data summary + filtering, (5) Performance benchmarks (MMLU + HumanEval + multiple), (6) Carbon footprint training estimation, (7) Citations + ethical use guidelines, comprehensive Model Card supports responsible AI deployment.

Termes liés

• Datasheets for Datasets — complementary documentation.