FrontierPhysics: 56 expert tasks benchmark AI agents on end-to-end physics research
Bingran You · hf · 2026-11-05
FrontierPhysics is a new end-to-end benchmark of 56 tasks designed by 42 physicists across nine theoretical, computational, and experimental domains. Agents must review literature, plan research, run experiments, and produce structured outputs and scientific manuscripts. Evaluation combines deterministic verifiers with expert-authored rubrics, judging both correctness and publishability. Paper on arXiv, with authors including Gerbrand Ceder.
More from coding & agent
- GitHub Next's PDD Kit uses formal methods to prove AI-generated web app code is correct — 0xkarasy · 2026-11-05
- Meta's JiTTesting deploys LLM-generated tests to catch bugs pre-merge, cutting human review 70% — AxSaucedo · 2026-11-05
- Built an app in a weekend with AI, now spending 3 weeks figuring out what it does — kylegawley · 2026-11-05
- Developer seeks open-source guardrails against malicious file uploads to LLM chatbots — SnooCupcakes982 · 2026-11-05
- Dev predicts 180Hz screen-frame models will make vibe coders obsolete within a year — ryunuck · 2026-11-05
- Agents ace isolated tasks but drop to ~38% at maintaining codebases over time — FullStackFrenzy · 2026-11-05