{
 "id": "WootzappLab/browser-agent-tasks",
 "owner": "WootzappLab",
 "name": "browser-agent-tasks",
 "downloads": 1037,
 "likes": 0,
 "last_modified": "2026-10-30T11:39:36.000Z",
 "stated_licence": "apache-2.0",
 "stated_from": "tag",
 "licence_name": null,
 "licence_link": null,
 "file_licence": null,
 "licence_files": [],
 "sources": [],
 "gated": false,
 "tasks": [],
 "citation": "WebVoyager introduced multimodal automatic evaluation for open-ended web\ntasks [[1]](#references). The screenshot baseline follows the Universal\nVerifier's rubric-centered design: criterion-specific evidence selection,\nseparate process and outcome judgments, and failure attribution\n[[2]](#references). WootzApp adds the DOM-based adaptation, browser-native\nevidence capture, paired comparison protocol, and source-aware audit.\n\nWebArena provides screenshot, HTML DOM, and accessibility-tree observations\nand evaluates functional task completion in controlled websites\n[[4]](#references). These systems provide context for structured and visual\nbrowser evidence. This benchmark measures what paired recorded\nrepresentations preserve under a shared rubric; it does not reproduce those\nbenchmarks' task-success experiments.",
 "signal": "permissive",
 "reason": "The declared licence (apache-2.0) is classified as permissive.",
 "revision": "7c961cc4ed9534151742ee9695504937074b195c",
 "stated_evidence": "https://huggingface.co/datasets/WootzappLab/browser-agent-tasks/blob/7c961cc4ed9534151742ee9695504937074b195c/README.md",
 "file_evidence": "https://huggingface.co/datasets/WootzappLab/browser-agent-tasks/tree/7c961cc4ed9534151742ee9695504937074b195c",
 "card_nc": false
}