Benchmark Almanac

Harvard

We track 5 benchmarks that Harvard made or helped make.

0 in use 4 with public code 3 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
FrontierPhysics Agents do end-to-end physics research on 56 tasks that 42 physicists from 26 institutions built from their own work, across nine domains of theoretical, computational and experimental physics. Each task asks the agent to review the literature, plan, run the research and write a manuscript. Deterministic verifiers and expert rubrics grade whether the research is correct and publishable, over 3 trials. Research science Aug 2026 27.4% Not in cards
TheoPhysEval Agents solve 50 unpublished research-level theoretical physics problems on their own, with literature search and scripts. Each final answer is checked against an expert answer, and the score is the pass rate. Research science Oct 2026 98% Not in cards
LIBERO-PRO LIBERO policies tested under object, position, instruction, task and environment perturbations. 50 episodes per task. Success rate. Simulated robots Oct 2025 53% Not in cards
AstroCLIMB Models compare pairs of astronomy figures or captions. They determine whether the pair shares a figure, shares a paper, comes from citation-linked papers, or comes from unrelated papers. Lab and data work Jul 2026 Not in cards
Detecting Entities in the Astrophysics Literature Models mark and classify named entities in excerpts from astrophysics papers, including telescopes, astronomical objects, and grants. Organizers score entity spans and token labels. Lab and data work May 2022 Not in cards