mrkeyoor.com_
Sat 08 Aug 21:02 UTC
AI Toolsevaluationupdated 08 Aug 2026

harvey-labs

Harvey LAB is an open benchmark for testing whether tool-using AI agents can complete realistic legal assignments. It packages synthetic matter files, detailed grading rubrics, an isolated execution harness, and reports so teams can compare agents on work such as contract review, drafting, research, and extraction.

Verdict

Harvey LAB is one of the strongest public starting points for testing legal agents because it evaluates produced files, not just short answers. Use it as an internal regression and comparison suite, after sampling tasks and rubrics for defects. Do not turn its all-pass rate into a procurement verdict without human review and repeated runs.

Setup3/5Good setup script, but containers, document tools, and API keys remain
Docs5/5Excellent tutorial, architecture map, methodology, and CLI reference
Community4/5Fast issue and pull request activity around tasks and harness code
Maturity3/5Version 1.0 is useful, but the young corpus is still being corrected

Who it’s for

  • Legal AI teams that need repeatable tests closer to work product than multiple-choice questions.
  • Model developers comparing agent behavior across providers on the same documents and rubrics.
  • Law-firm innovation groups willing to audit benchmark tasks before using results in procurement.
  • Researchers studying document coverage, tool use, cost, latency, and criterion-level failures.

Who it’s NOT for

  • Teams seeking a benchmark made from pristine lawyer-authored matters. The tutorial says documents are synthetically generated in batches and contain imperfections.
  • Buyers who need one stable, objective leaderboard number. Grading uses LLM judges, and open issue 114 reports provider-dependent behavior in one scoring path.
  • Evaluators who want partial task credit as the headline result. One failed criterion makes the entire task score zero under LAB's all-pass method.
  • Anyone planning to substitute benchmark success for attorney review. The tasks imitate legal work, but neither synthetic files nor judge verdicts establish safe client-facing performance.

Setup reality

The setup script is a credible shortcut on a supported machine, but a complete run still needs Python 3.12, uv, Podman, document tooling, and at least an Anthropic API key for the default judge. The tutorial estimates about 20 minutes for one small agent run and judge pass, most of it spent on model calls. Windows setup requires Windows 11, virtualization, WSL2 installation, and an initial reboot. A single task is approachable; broad sweeps across 1,600-plus tasks quickly become an API-cost, storage, and results-management exercise.

Legal evaluation that looks like work

Most legal AI benchmarks ask a model to classify a passage, recall a rule, or choose an answer. Harvey LAB asks an agent to work through a matter file and produce the artifact a lawyer requested. A corporate M&A task might provide 60 synthetic documents and expect a Word memorandum. Another task may require extracting terms into a structured file, reviewing an agreement, or drafting a package. That change in unit of evaluation makes LAB much more relevant to teams building agents that read and write real office documents.

The repository combines 1,600-plus tasks across 24 legal practice areas and contracting with a Python execution harness. Each task supplies instructions, source documents, named deliverables, and expert-written pass or fail criteria. The project is MIT licensed, tagged v1.0, and includes no database or hosted service. Tasks live on disk, generated runs go under results/, and reports are static HTML. This simple layout makes runs inspectable and easy to archive.

The harness is the real differentiator

LAB does not stop at publishing a dataset. Its harness places an agent in a per-task Podman container with networking disabled, Linux capabilities dropped, source documents mounted read-only, and a writable output area. Six tools let the agent run shell commands, read, write, edit, search, and list files. Parsing support covers Word, Excel, PowerPoint, PDF, and text through tools including Pandoc, MarkItDown, pandas, and pdfplumber.

That environment gives evaluations a consistent operational shape. The harness records the full transcript, token use, wall time, files written, and how many documents the agent read or skipped. Teams can inspect a failure as a trajectory, not merely a bad final answer. Adapters cover Anthropic, OpenAI, Google, Mistral, and Fireworks model routes, so the same task can exercise different providers without rewriting its documents or rubric.

The tutorial is excellent. It walks through environment setup, model configuration, task inspection, a run, grading, report reading, alternative models, sweeps, and comparison dashboards. A first small run is estimated at about 20 minutes. The example is concrete enough that an evaluation engineer should reach a useful report without reverse-engineering the package.

Strict scoring is useful and easy to misuse

Every rubric criterion receives a binary verdict from an LLM judge. A task earns 1.0 only if every criterion passes; one miss produces 0.0. LAB still reports the fraction of criteria passed as a diagnostic. Its rationale is sensible for legal work: a diligence memo that finds most issues but omits one material problem can still fail its purpose.

The problem is that strictness magnifies errors in the benchmark itself. Open issue 115 documented a rubric that mixed merger-guideline thresholds, causing a correct answer to fail the entire task. Recent pull requests have corrected stale entity names, source-matter mismatches, and rubric details. The maintainers are responding quickly, but users should treat task quality as something to audit, not an assumption inherited from the v1.0 tag.

LLM judging creates another layer of uncertainty. The default judge evaluates each criterion semantically against the relevant deliverable, with no golden answer. Optional dual judging runs two named judges and preserves both outputs. That is better than hiding disagreement, yet it does not make verdicts objective. Open issue 114 reports provider-dependent behavior around fuzzy deliverable matching. Serious comparisons should repeat runs, preserve judge versions, inspect borderline reasoning, and avoid ranking models from a single sweep.

Setup is manageable, running it is not free

On macOS or Linux, cloning the repo and running scripts/setup.sh handles much of the initial environment. The project requires Python 3.12, uses uv, and depends on document parsers plus provider SDKs. Podman supplies isolation. Windows users need Windows 11, enabled CPU virtualization, WSL2, and a reboot on the first setup.

A model-provider key is required for the agent, and the default judge needs an Anthropic key. Dual judging also needs an OpenAI key. A single task is easy to budget; a practice-area sweep can multiply agent turns, criterion-level judge calls, output files, and reports rapidly. The comparison tools report latency and estimated cost, which helps, but teams should dry-run a narrow selection before pointing the harness at the full corpus.

Use it as a test suite, not a certification

Project health is strong for such a young repository. Harvey LAB was created in March 2026, had about 650 stars at review time, and was pushed on August 7. GitHub showed 46 open issues and pull requests combined, with new fixes and task corrections active on August 8. There is a v1.0 tag but no GitHub Release entry, so current push and issue activity are better health signals than release cadence.

The right use is an internal regression suite. Select tasks matching your product, have lawyers inspect their source files and criteria, pin the repository revision and judge configuration, then compare repeated runs. LAB offers rare visibility into document handling and missed requirements. It cannot prove that an agent is ready for client work, but it can expose where that agent fails before clients do.

Alternatives

ProjectWhat it isPick it when
LegalBenchAn open collection of legal reasoning tasks for evaluating foundation models.Pick this instead when you need narrower legal reasoning tests without running a file-editing agent sandbox.
HELMA framework for transparent, multi-metric evaluation of language and multimodal models.Pick this instead when broad model evaluation and standardized scenarios matter more than legal work products.
LM Evaluation HarnessA widely used framework for reproducible few-shot language-model evaluation.Pick this instead when academic task coverage and lightweight model scoring are the priority.

What people are saying

  1. [github-trending] harveyai/harvey-labs

Sources

  1. Harvey LAB repository
  2. Harvey LAB tutorial
  3. Harvey LAB architecture
  4. Harvey LAB evaluation methodology
  5. Provider-dependent scoring issue
  6. Incorrect rubric threshold issue