FrontierPhysics: Evaluating Agents for End-to-End Physics Research
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
FrontierPhysics is a benchmark of 56 end-to-end physics research tasks, written by 42 physicists from their own projects across nine domains. Agents must review literature, plan, run experiments, and produce a manuscript, graded by deterministic verifiers and expert rubrics.