Harvey Legal Agent Benchmark
Over 1,200 law firm agent tasks across 24 practice areas, scored against more than 75,000 expert-written rubric criteria.
Description
Harvey open-sourced its Legal Agent Benchmark on May 6, 2026. It has over 1,200 agent tasks across 24 practice areas, scored against more than 75,000 expert-written rubric criteria.
The tasks are law firm work. The method used to build the rubric criteria is relevant to anyone writing evaluation protocols for legal help tools.
Access the Dataset
https://github.com/harveyai/harvey-labsJusticeBench lists this dataset for research and development purposes only. Listing it does not constitute an endorsement, recommendation, or approval by the Stanford Legal Design Lab. See the Terms of Use.