Stanford Center for Research on Foundation Models (CRFM)
- Type
- Academic institution
- Lieu
- États-Unis — United States (Stanford, within HAI)
- Dernière vérification
- 2026-09-22
- Prochaine vérification
- 2027-03-21
Pas encore traduit — affiché en anglais.
Comment les joindre
- HELM — open benchmark; models and scenarios can be contributed via GitHub
- Foundation Model Transparency Index — developers respond to it, and the methodology is public
Ce canal est conservé uniquement comme référence. Sa vérification est absente, date de plus de 180 jours, ou la période de contact n'est pas ouverte actuellement ; ne vous y fiez pas avant une nouvelle vérification.
- People directory — the practical contact route is an individual researcher
- Homepage
Ce qu'il fait
Interdisciplinary centre spanning 10+ Stanford departments. Coined the term 'foundation models'. Outputs include HELM (Holistic Evaluation of Language Models — a widely used open benchmark suite), the Foundation Model Transparency Index (FMTI, which grades developers on transparency), Ecosystem Graphs, and Levanter. Works across technical foundations, beneficial applications, societal impact (transparency, openness, copyright, privacy, systemic risk) and AI policy.
Évaluation franche
No centre-level inbox at all. Everything runs through named researchers or through open-source contribution to HELM, which is the genuinely open channel.
Préoccupations couvertes
Systemic risk from homogenisation and opacity — the founding thesis is that a small number of foundation models become single points of failure across society. Transparency-centred. Covers near-term and structural risk; touches systemic risk but not loss of control.
Canal gouvernemental
Real — CRFM has filed influential comments in US and EU processes (notably on the EU AI Act and on open foundation models) and its transparency work is cited by regulators.