{"record": {"aliases": ["REDAgentBench: Executable Red Teaming and Faithful Measurement of LLM Agent Systems"], "artifacts": [{"id": "arxiv:2608.10669", "kind": "paper", "url": "https://arxiv.org/abs/2608.10669"}], "categories": ["Agents", "Agents & Tool Use", "Cybersecurity"], "description": {"en": "Evaluates LLM agent safety through executable red-teaming, adversarial case generation, and verification of harmful effects across 1,661 cases and five service surfaces."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.10669", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.10669", "title": "REDAgentBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.10669", "modality": null, "name": "REDAgentBench", "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_redagentbench-executable-red-teaming-and-f_69dd1033", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "fbe3fd5c2d94f3b6e1c66a11b80660a9400be372166094a359413344221f7449", "relation": "evaluates_only", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-08-27T04:14:52.836173Z", "source_url": "https://arxiv.org/abs/2608.10669"}, "publisher": null, "released": "2026-08-11", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.10669", "source_url": "https://arxiv.org/abs/2608.10669"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-10669", "source": "claire_radar", "source_benchmark_id": "2608.10669", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["REDAgentBench: Executable Red Teaming and Faithful Measurement of LLM Agent Systems"], "annotation": "Unknown", "applicationDomains": ["Cybersecurity"], "area": "Agents & Tool Use", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2608.10669", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Agent safety and red-teaming are high-interest topics, and the benchmark's executable evaluation and measurement innovations should attract attention in the LLM safety community.", "score": 58}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:14:52.836173Z", "inputHash": "c4fa39415a4e17cf38a6b40331ca5e49368153a40cff61f2bb17ebbc93fc2c87", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce REDAgentBench, an executable framework for autonomous red-teaming and faithful measurement", "canonicalNameSource": "abstract", "decisionReason": "The benchmark is explicitly named, defines a reproducible executable evaluation with verification, and includes a public measurement framework.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:14:52.836173Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates LLM agent safety through executable red-teaming, adversarial case generation, and verification of harmful effects across 1,661 cases and five service surfaces.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "evaluation protocol evidence", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We introduce REDAgentBench, an executable framework for autonomous red-teaming and faithful measurement."}, "familyId": "bmf_8ee636d10fcd", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_redagentbench-executable-red-teaming-and-f_69dd1033", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Cybersecurity"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2608.10669", "pdf": "https://arxiv.org/pdf/2608.10669", "project": null, "report": "https://arxiv.org/abs/2608.10669"}, "metrics": [], "motivation": "Large language model (LLM) agents combine language-based reasoning with external tools to perform complex tasks.", "name": "REDAgentBench", "oneLine": "Evaluates LLM agent safety through executable red-teaming, adversarial case generation, and verification of harmful effects across 1,661 cases and five service surfaces.", "paperTitle": "REDAgentBench: Executable Red Teaming and Faithful Measurement of LLM Agent Systems", "primaryDomain": "Cybersecurity", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.10669", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "ranking": {"90d": {"confidence": "Low", "coverage": 0.15, "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": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.0, "value": 0}}, "confidence": "Low", "coverage": 0.15, "rank": null, "score": 43}, "method": "attention-ranking-v13", "rank": null, "score": 43}}, "readiness": "Paper only", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-08-11", "source": {"authors": ["Zixing Chen", "Xingyuan Liu", "Jie Zhu", "Huaixia Dou", "Shuo Jiang", "Junhui Li", "Lifan Guo", "Feng Chen", "Chi Zhang"], "categories": ["cs.AI"], "id": "2608.10669", "title": "REDAgentBench: Executable Red Teaming and Faithful Measurement of LLM Agent Systems", "type": "arxiv", "url": "https://arxiv.org/abs/2608.10669"}, "sourceUpdatedAt": "2026-08-11T08:48:54Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Provides an executable and measurable approach for agent safety evaluation beyond aggregate attack success rates."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
