{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.22510", "kind": "paper", "url": "https://arxiv.org/abs/2608.22510v1"}, {"id": "hf:xyh110sym/clawprobench", "kind": "dataset", "url": "https://huggingface.co/datasets/xyh110sym/clawprobench"}], "authors": ["YuanHang Xiao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.22510", "source_url": "https://arxiv.org/abs/2608.22510"}, "categories": ["Agents", "General AI", "Language & Knowledge"], "description": {"en": "Evaluates agent configurations on 102 live-runtime and 68 frozen holdout scenarios, scoring execution traces with a safety-gated formula covering correctness, process quality, and efficiency."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.22510v1", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.22510v1", "title": "ClawProBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Chinese University of Hong Kong"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.22510", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.22510", "source_url": "https://openalex.org/W7204191359"}, "key": "claire-radar:2608.22510", "modality": null, "name": "ClawProBench", "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": "true", "origin_record_id": "bm_clawprobench_7f331d78", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "7f6841b16ce7c0e10b534b767ecd80fa098f6744acd8f74af92821c8cabfa78f", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "source-reviewed", "reviewed_at": "2026-08-30", "source_url": "https://arxiv.org/abs/2608.22510v1"}, "publisher": null, "released": "2026-08-23", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.22510", "source_url": "https://arxiv.org/abs/2608.22510v1"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-22510", "source": "claire_radar", "source_benchmark_id": "2608.22510", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-08-26", "attemptedAt": "2026-08-25T17:51:33.814358Z", "githubRepo": "https://github.com/suyoumo/ClawProBench", "githubScope": "benchmark_repo", "githubStars": 822, "hfDailySubmittedAt": "2026-08-25T00:00:00.000Z", "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 2, "hfPaperUrl": "https://huggingface.co/papers/2608.22510", "observedAt": "2026-08-25T17:51:33.814358Z", "signalStatus": {"githubStars": {"observedAt": "2026-08-25T17:51:33.814358Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-08-25T17:51:33.814358Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "A trace-aware agent benchmark with a frozen holdout and multi-configuration results addresses a current gap and should attract research community attention.", "score": 72}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:18:56.318436Z", "inputHash": "ccfe8c9bfb02e81e45230688dc1230abfdfede0e45b8b2324e6d6c829b3b0fd0", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We present ClawProBench, a trace-aware benchmark for runtime-native agent evaluation instantiated on OpenClaw", "canonicalNameSource": "abstract", "decisionReason": "The paper defines two scored tracks, safety-gated scoring, and evaluation manifests with sanitized traces, providing a credible comparison path.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-30", "sources": ["https://arxiv.org/abs/2608.22510", "https://github.com/suyoumo/ClawProBench", "https://huggingface.co/datasets/xyh110sym/clawprobench"], "state": "source-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates agent configurations on 102 live-runtime and 68 frozen holdout scenarios, scoring execution traces with a safety-gated formula covering correctness, process quality, and efficiency.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We present ClawProBench, a trace-aware benchmark for runtime-native agent evaluation instantiated on OpenClaw, a live agent runtime with workspace tools and native surfaces for browsing, memory, messaging, scheduling, skills, and subagents."}, "familyId": "bmf_dbc550709e63", "firstSeenAt": "2026-08-25", "heat": null, "id": "bm_clawprobench_7f331d78", "indexedAt": "2026-08-25T17:37:25.889905Z", "industrySectors": [], "links": {"code": null, "data": "https://huggingface.co/datasets/xyh110sym/clawprobench", "paper": "http://arxiv.org/abs/2608.22510v1", "pdf": "https://arxiv.org/pdf/2608.22510v1", "project": null, "report": "http://arxiv.org/abs/2608.22510v1"}, "metrics": [], "motivation": "Agent benchmarks often evaluate only final answers even when agents run on stateful runtimes.", "name": "ClawProBench", "oneLine": "Evaluates agent configurations on 102 live-runtime and 68 frozen holdout scenarios, scoring execution traces with a safety-gated formula covering correctness, process quality, and efficiency.", "paperTitle": "ClawProBench: Trace-Aware Evaluation of AI Agents with Runtime Coverage and Frozen Workplace-Style Holdouts", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "http://arxiv.org/abs/2608.22510v1", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-25T17:37:25.889905Z"}, "publications": [], "ranking": {"30d": {"confidence": "High", "coverage": 0.85, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"percentile": 0.996268656716418, "value": 822}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": 0.44554455445544555, "value": 2}}, "confidence": "High", "coverage": 0.85, "rank": 17, "score": 76}, "method": "attention-ranking-v11", "rank": 17, "score": 76}, "90d": {"confidence": "Medium", "coverage": 0.7, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"percentile": 0.9930555555555556, "value": 822}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": 0.5580568720379147, "value": 2}}, "confidence": "Medium", "coverage": 0.7, "rank": 26, "score": 78}, "method": "attention-ranking-v11", "rank": 26, "score": 78}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-23", "source": {"authors": ["YuanHang Xiao"], "categories": ["cs.AI"], "id": "2608.22510", "title": "ClawProBench: Trace-Aware Evaluation of AI Agents with Runtime Coverage and Frozen Workplace-Style Holdouts", "type": "arxiv", "url": "http://arxiv.org/abs/2608.22510v1"}, "sourceUpdatedAt": "2026-08-23T17:09:02Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "It moves agent evaluation beyond final answers to expose runtime-navigation, safety-boundary, and repeated-execution failures that conventional leaderboards hide."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
