Back
Stanford CRFM
webcrfm.stanford.edu·crfm.stanford.edu/
CRFM is a leading academic institution for foundation model research; relevant to AI safety for its evaluation frameworks (HELM), governance work, and foundational analysis of large-scale AI risks and societal impacts.
Metadata
Importance: 62/100homepage
Summary
CRFM at Stanford University is an interdisciplinary research center focused on the study, development, and responsible deployment of foundation models (large-scale AI models like GPT and BERT). It conducts technical research, policy analysis, and evaluation of foundation models while addressing their societal implications. CRFM is known for producing influential work including the original 'foundation models' terminology and the HELM benchmark.
Key Points
- •Coined the term 'foundation models' and published the seminal 2021 report on their capabilities, risks, and societal impact
- •Develops HELM (Holistic Evaluation of Language Models), a comprehensive benchmark for evaluating LLM performance across diverse tasks
- •Conducts interdisciplinary research spanning technical AI, law, ethics, economics, and policy related to large-scale AI systems
- •Produces open resources including model evaluations, datasets, and research on responsible AI deployment
- •Part of Stanford HAI (Human-Centered AI), bridging academic research and real-world AI governance concerns
1 FactBase fact citing this source
Cached Content Preview
HTTP 200Fetched Apr 5, 20262 KB
Stanford CRFM About Us The Center for Research on Foundation Models (CRFM) is an interdisciplinary initiative at the Stanford Institute for Human-Centered Artificial Intelligence (HAI) that makes fundamental advances in the study, development, and deployment of foundation models. The center brings together a multidisciplinary group of faculty, students, researchers, and engineers spanning 10+ departments at Stanford. CRFM initiatives relate to: Technical foundations. We conduct research to better understand and build better foundation models: our research spans data, systems, architectures, training, adaptation, inference, interpretability, evaluation, and new technical paradigms. Beneficial applications. We build foundation models and other resources in specialized domains like law, music, robotics, and biomedicine. Societal impact. We conceptualize and clarify how foundation models impact society by considering matters like transparency, supply chains, openness, copyright, privacy, and systemic risk. AI policy. We support governments in formulating and implementing evidence-based policy across multiple jurisdictions. What is a foundation model? Foundation models are powerful resource-intensive models that have emerged as the dominant paradigm for AI in the 2020s. To learn more, watch our explainer . WHY DO WE CARE? Foundation models present novel capabilities and risks that are poorly understood. Nonetheless, they are being deployed at scale.
Resource ID:
kb-02ad63a54fdb1d8b | Stable ID: sid_c9bLQwxsWN