{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.05573", "kind": "paper", "url": "https://arxiv.org/abs/2608.05573"}, {"id": "gh:hanzhi306/skilltv-bench", "kind": "repo", "url": "https://github.com/HanZhi306/SkillTV-Bench"}], "authors": ["Zhi Han", "Chenxi Zeng", "Liuhaichen Yang", "Zihan Guo", "Ming Zhou", "Yang Li"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.05573", "source_url": "https://arxiv.org/abs/2608.05573"}, "categories": ["Agents", "General AI", "Language & Knowledge"], "description": {"en": "Evaluates whether a judge can verify long-horizon agent executions using original instructions, normalized trajectories, task-time skills, and inspectable artifacts in a runnable environment."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.05573", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.05573", "title": "SkillTV-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Shanghai Jiao Tong University", "Shanghai Artificial Intelligence Laboratory", "University College London", "Sun Yat-sen University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.05573", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.05573", "source_url": "https://openalex.org/W7201881355"}, "key": "claire-radar:2608.05573", "modality": null, "name": "SkillTV-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_skilltv-bench_a205ca2b", "origin_source": "arxiv", "recognition_confidence": "0.75", "record_sha256": "bc3e8f0eef4df413e6046404f500124c00029eeb81a70e5ef566d49840b52516", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-06T08:45:23.913416Z", "source_url": "https://arxiv.org/abs/2608.05573"}, "publisher": null, "released": "2026-08-06", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.05573", "source_url": "https://arxiv.org/abs/2608.05573"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-05573", "source": "claire_radar", "source_benchmark_id": "2608.05573", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/HanZhi306/SkillTV-Bench", "githubScope": "benchmark_repo", "githubStars": 4, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.05573", "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", "state": "unavailable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark addresses agentic evaluation with released code and data, a clear scoring protocol, and a proposed skill evolution method that may draw interest from the agent LLM community.", "score": 68}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-06T08:45:23.913416Z", "inputHash": "d61c239fd79f284fd72c8bd6acf3c28c2335f9e2c2cf25b1ecbad8e7074272fe", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "# SkillTV-Bench: Benchmarking How Well Judges Perform on Skill-Augmented Agentic Execution", "canonicalNameSource": "official_readme", "decisionReason": "The benchmark provides a fixed 203-case evaluation split with a Harbor verifier that hides labels and allows candidate gating, along with public code and data on GitHub enabling reuse.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-06T08:45:23.913416Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates whether a judge can verify long-horizon agent executions using original instructions, normalized trajectories, task-time skills, and inspectable artifacts in a runnable environment.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "evaluation protocol evidence", "public artifact URL"], "snippet": "We therefore introduce SkillTV-Bench, a 681-case benchmark of real agent trajectories from 50 tasks across eleven domains, designed to evaluate skill-aware trajectory verification for both LLM-as-a-Judge and Agent-as-a-Judge methods."}, "familyId": "bmf_138d1f2f0f96", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_skilltv-bench_a205ca2b", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/HanZhi306/SkillTV-Bench", "data": null, "paper": "https://arxiv.org/abs/2608.05573", "pdf": "https://arxiv.org/pdf/2608.05573", "project": null, "report": "https://arxiv.org/abs/2608.05573"}, "metrics": [], "motivation": "LLM agents increasingly execute long-horizon tasks through tool use and environment interaction, shifting evaluation from final-response scoring to verification of complete executions.", "name": "SkillTV-Bench", "oneLine": "Evaluates whether a judge can verify long-horizon agent executions using original instructions, normalized trajectories, task-time skills, and inspectable artifacts in a runnable environment.", "paperTitle": "SkillTV-Bench: Benchmarking How Well Judges Perform on Skill-Augmented Agentic Execution", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.05573", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "HanZhi306", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/HanZhi306/SkillTV-Bench"}], "ranking": {"90d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.21483102985913266, "value": 4}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 202, "score": 34}, "method": "attention-ranking-v13", "rank": 202, "score": 34}}, "readiness": "Runnable", "recognitionConfidence": 0.75, "relation": "introduces", "releasedAt": "2026-08-06", "source": {"authors": ["Zhi Han", "Chenxi Zeng", "Liuhaichen Yang", "Zihan Guo", "Ming Zhou", "Yang Li"], "categories": ["cs.AI"], "id": "2608.05573", "title": "SkillTV-Bench: Benchmarking How Well Judges Perform on Skill-Augmented Agentic Execution", "type": "arxiv", "url": "https://arxiv.org/abs/2608.05573"}, "sourceUpdatedAt": "2026-08-06T03:52:37Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Existing judge benchmarks rarely expose task-time procedural knowledge and inspectable artifacts together, limiting evaluation of evidence-grounded verdicts for skill-augmented agents."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
