About

This is an independent, open-source visualization of Harvey AI's harvey-labs legal agent benchmark. The goal is to make it easy to see what kinds of legal work the benchmark covers and to look at the actual tasks, rubrics, and reference materials that an agent is graded against.

Every task page reads its data from the upstream repository pinned at a specific commit, so links and contents stay stable between data refreshes. Reference documents are not redistributed here. Each file links back to its original location on GitHub.

Data provenance

• Source: harveyai/harvey-labs@ccafaf51
• License of upstream data: MIT (see the upstream LICENSE)
• Last data refresh: 2026-05-11

How this site is built

At build time, a script walks the vendored copy of the benchmark, validates every task.json against a Zod schema, and emits one static page per task plus an index for listings and search. The interactive bits (rubric collapse, list filters) ship as small React islands inside an otherwise zero-JavaScript page.

Attribution

This site is not affiliated with, endorsed by, or sponsored by Harvey AI. "Harvey", "BigLaw Bench", and related marks belong to their respective owners.