


Sanmi Koyejo
Researcher
Now Meta
Now Assistant Professor of Computer Science · Stanford University
@sanmikoyejocs.stanford.eduLinkedInLatest move
Virtue AI → Meta
Path
Current
- MetaNEWJoined
Previous
Virtue AI– 2026Co-founder
Stanford University (stai_research group)Leads the lab
Papers
- DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT ModelsarXiv 2023Sets out a benchmark that tests GPT models on toxicity, bias, robustness, privacy, fairness and related trust criteria.
- Are Emergent Abilities of Large Language Models a Mirage?arXiv 2023Argues that apparent sudden jumps in language model abilities can come from the choice of evaluation metric.
Interviews and talks
- 2026-08-20Sanmi Koyejo: Stanford Professor '50% of AI Is Data — 10% Is the Algorithm We Actually Teach' |
Ep 9 · Practice Ground · Podcast ↗ - 2025-11-28Ep. 31: Sanmi Koyejo, Stanford CS Professor and Virtue
AI Co-Founder · Delta Podcast · Podcast ↗ - 2024-02-12Are Emergent Behaviors in LLMs an Illusion? with
Sanmi Koyejo · The TWIMLAI Podcast · Podcast ↗ - 2020-02-27Metric Elicitation and Robust Distributed Learning with
Sanmi Koyejo · The TWIMLAI Podcast · Podcast ↗
On the timeline
SourcesLatest move: semafor.comapi.fxtwitter.comcs.stanford.edu
Something wrong or missing on this page? Tell me and it will be fixed or removed. Contact →