Benchmark

LegalBench

Measures: 162 legal reasoning tasks (issue spotting, rule recall, conclusion, interpretation, rhetoric) built by legal professionals, mostly classification or short answer, scored by accuracy.

Maintainer
Stanford HazyResearch / LegalBench contributors (Guha et al.)
Introduced
2023
Status
Active
Saturation
Saturated
Typical provenance
Independent
In scoring set
No

Paper

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

No results to show.

LegalBench: leaderboard and what it measures · supermarket.dev