Websites · PROJECT NOTES
Mind2Web 2
An agentic-search benchmark for long-horizon web research and citation-backed synthesis, with a released judge-agent framework and versioned task-specific evaluators.
What can you do with it?
Assess completeness and source attribution of synthesized answers with task-specific verification trees and cached evidence pages.
What should you know before using it?
Distinct from offline Mind2Web action prediction and Online-Mind2Web navigation. Public Git code includes 10 development evaluators; current full task/evaluator files on Hub are auto-gated and were not accessed. Full evaluator release was announced in October 2025, so the paper-era private-test description is historical. Live pages and model judgments require version/date controls.
What is verified here?
Runtime: Python >=3.11, browser/webpage-cache services and judge models; some tasks use Google Maps APIs. The runner defaults to 2025_07_14 evaluators while Hub lists 2025_10_23 scripts. Align script release, task split and caches. No browser or evaluator was executed.
License note: Git repository code is MIT. Public Hub metadata declares CC BY 4.0 for the gated task/evaluator distribution. Live source pages retain their own terms. Upstream license ↗
This entry was reviewed against original project material. We have not independently run the software or reproduced its results. AI-development claims are author-reported; check upstream for current versions, setup requirements, and licenses.
- Resource type
- Benchmark suite
- Provenance
- Development provenance not established
- Reviewed
- Platforms
- Web