{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.02404", "kind": "paper", "url": "https://arxiv.org/abs/2606.02404"}, {"id": "gh:prometheus-eval/k-browsecomp", "kind": "repo", "url": "https://github.com/prometheus-eval/K-BrowseComp"}], "authors": ["Nahyun Lee", "Dongkeun Yoon", "Guijin Son", "Geewook Kim", "Dayoon Ko", "Jeonghun Park", "Haneul Yoo", "Jaewon Cho", "Junghun Park", "Changyoon Lee", "Kyochul Jang", "Jaeyeon Kim", "Eunsu Kim", "Woojin Cho", "Seungone Kim"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.02404", "source_url": "https://arxiv.org/abs/2606.02404"}, "categories": ["Agents", "Cybersecurity", "Language & Knowledge"], "description": {"en": "K-BrowseComp evaluates web-browsing agents on 400 Korean-context problems, requiring multi-hop or parallel evidence retrieval from public Korean websites and returning a single short answer. Includes a 300-problem manually verified subset and a 100-problem synthetic stress-test split."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.02404", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.02404", "title": "K-BrowseComp"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.02404", "modality": null, "name": "K-BrowseComp", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_k-browsecomp_c3715567", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "7a074d15e23db6fcbab6faf6edd43bf6fadcfd98a80631fa0b17d67b5ec61829", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.02404"}, "publisher": null, "released": "2026-06-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.02404", "source_url": "https://arxiv.org/abs/2606.02404"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-02404", "source": "claire_radar", "source_benchmark_id": "2606.02404", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Machine Generated", "applicationDomains": ["Cybersecurity"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/prometheus-eval/K-BrowseComp", "githubScope": "benchmark_repo", "githubStars": 16, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 57, "hfPaperUrl": "https://huggingface.co/papers/2606.02404", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "5809d0ceaf4b49ab9433161381ed8445813f7968c9c9ca3a4c74fd5299c03617", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "K-BrowseComp evaluates web-browsing agents on 400 Korean-context problems, requiring multi-hop or parallel evidence retrieval from public Korean websites and returning a single short answer. Includes a 300-problem manually verified subset and a 100-problem synthetic stress-test split.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence"], "snippet": "We introduce K-BrowseComp, a web-browsing agent benchmark grounded in Korean contexts, consisting of 400 problems."}, "familyId": "bmf_7a5a02770094", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/prometheus-eval/K-BrowseComp", "url": "https://github.com/prometheus-eval/K-BrowseComp"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/prometheus-eval/K-BrowseComp", "url": "https://huggingface.co/papers/2606.02404"}], "scope": "benchmark_repo", "url": "https://github.com/prometheus-eval/K-BrowseComp", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_k-browsecomp_c3715567", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Cybersecurity"], "links": {"code": "https://github.com/prometheus-eval/K-BrowseComp", "data": null, "hfPaper": "https://huggingface.co/papers/2606.02404", "paper": "https://arxiv.org/abs/2606.02404", "pdf": "https://arxiv.org/pdf/2606.02404", "project": null, "report": "https://arxiv.org/abs/2606.02404"}, "metrics": [], "motivation": "Frontier model evaluations are shifting from foundational capabilities (e.g., instruction following and reasoning) toward compositional, agentic ones, but Korean agentic benchmarks remain scarce.", "name": "K-BrowseComp", "oneLine": "K-BrowseComp evaluates web-browsing agents on 400 Korean-context problems, requiring multi-hop or parallel evidence retrieval from public Korean websites and returning a single short answer. Includes a 300-problem manually verified subset and a 100-problem synthetic stress-test split.", "paperTitle": "K-BrowseComp: A Web Browsing Agent Benchmark Grounded in Korean Contexts", "primaryDomain": "Cybersecurity", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.02404", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "prometheus-eval", "organizationType": "community", "role": "benchmark-publisher", "sourceUrl": "https://github.com/prometheus-eval/K-BrowseComp"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-01", "source": {"authors": ["Nahyun Lee", "Dongkeun Yoon", "Guijin Son", "Geewook Kim", "Dayoon Ko", "Jeonghun Park", "Haneul Yoo", "Jaewon Cho", "Junghun Park", "Changyoon Lee", "Kyochul Jang", "Jaeyeon Kim", "Eunsu Kim", "Woojin Cho", "Seungone Kim"], "categories": ["cs.CL"], "id": "2606.02404", "title": "K-BrowseComp: A Web Browsing Agent Benchmark Grounded in Korean Contexts", "type": "arxiv", "url": "https://arxiv.org/abs/2606.02404"}, "sourceUpdatedAt": "2026-06-01T15:50:03Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Existing agentic benchmarks overlook Korean-language browsing, and frontier models show a significant performance drop on this benchmark. It provides a public protocol for measuring Korean web navigation and evidence-tracking abilities, which is relevant for deploying agents in Korean-language settings."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
