person · draft
Percy Liang
Stanford computer scientist, co-author of the generative agents work that simulates people with language models, and a co-founder of Simile.
Percy Liang teaches computer science at Stanford University and founded its Center for Research on Foundation Models, which he directs.1,2 He studied at MIT and took his PhD at UC Berkeley in 2011, advised by Michael Jordan and Dan Klein.1,2 Simile credits him with creating SQuAD and HELM.3 His CodaLab Worksheets platform keeps the full provenance of an experiment from raw data to final results.2 His current project, Marin, builds language models in the open, with experiments preregistered and visible as they run.2
He co-advised Joon Sung Park's doctorate with Michael Bernstein and is an author of the papers that built simulated people from language models.2 He is a co-founder and chief scientist of Simile, which sells those simulations.3
On the wiki he is an author of Generative Agents, of Generative Agent Simulations of 1,000 People and of Whose Opinions Do Language Models Reflect?, which measured how far language models' answers sit from those of real groups of Americans. Simile names him among its founders.
- affiliation
- Stanford University
- role
- computer scientist
- homepage
- https://cs.stanford.edu/~pliang/
makes
-
Stanford-led study, 2024, revised 2026: agents built from two-hour interviews with 1,052 Americans predicted their survey answers at 83% of the people's own two-week retest consistency.
open sourcerestricted data
-
ICML 2023 paper that compared nine language models with 1,498 Pew survey questions answered by 60 US demographic groups and found the models' opinions far from the public's.
open data
-
Research code from Stanford for language-model characters that remember, reflect and plan, shown in Smallville, a small game town of 25 agents; a 2024 follow-up repository builds agents from interviews and survey answers and lets researchers put questions to them.
open sourcepartly open dataopen participation
-
Company founded by the Stanford researchers behind generative agents that sells simulations of how customers, employees or populations would respond to a change, from models it says are trained on data about real people.
closed sourceclosed data
Tags: foundation models, machine learning, generative agents