{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.05462", "kind": "paper", "url": "https://arxiv.org/abs/2607.05462"}], "authors": ["Edwin H. Wintermute", "Harmon Bhasin", "Christina M. Agapakis", "Dianzhuo Wang", "Evan Seeyave", "Arjun Banerjee", "Daniel Fulop", "Matthew C. Watson", "Adam J. Meyer", "Sandrine Boissel", "Jens H. Kuhn", "Rishi Jain", "Noah D. Taylor", "Helena Shomar", "Patrick M. Boyle", "Kenny Workman"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.05462", "source_url": "https://arxiv.org/abs/2607.05462"}, "categories": ["General AI", "Language & Knowledge", "Safety"], "description": {"en": "BioSecBench-Refusal evaluates AI agents on 61 Routine and 46 Red-Team biosecurity-related tasks, measuring refusal rates and risk identification across multiple model configurations."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.05462", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.05462", "title": "BioSecBench-Refusal"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.05462", "modality": null, "name": "BioSecBench-Refusal", "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_biosecbench-refusal_d798db06", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "419d3733b66d8d1b413c7bc5826f5e5193d3092a3e89557a5894a496cda31312", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.05462"}, "publisher": null, "released": "2026-07-06", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.05462", "source_url": "https://arxiv.org/abs/2607.05462"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-05462", "source": "claire_radar", "source_benchmark_id": "2607.05462", "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": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.05462", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "71677a53bbce6ee720e8803b2e8c1cf8f83cbb421da9d89c5b1b99e1e5d8bf29", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "BioSecBench-Refusal evaluates AI agents on 61 Routine and 46 Red-Team biosecurity-related tasks, measuring refusal rates and risk identification across multiple model configurations.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We present BioSecBench-Refusal, a benchmark for risk identification and refusal behavior for biological research tasks."}, "familyId": "bmf_f88a84efd022", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_biosecbench-refusal_d798db06", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2607.05462", "paper": "https://arxiv.org/abs/2607.05462", "pdf": "https://arxiv.org/pdf/2607.05462", "project": null, "report": "https://arxiv.org/abs/2607.05462"}, "metrics": [], "motivation": "As AI agents are incorporated into life science workflows, the capabilities that speed discovery might also enable misuse.", "name": "BioSecBench-Refusal", "oneLine": "BioSecBench-Refusal evaluates AI agents on 61 Routine and 46 Red-Team biosecurity-related tasks, measuring refusal rates and risk identification across multiple model configurations.", "paperTitle": "BioSecBench-Refusal: A paired metric for performance and alignment in agentic biosecurity risk assessment", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.05462", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "arXiv", "organizationType": "benchmark-organization", "role": "benchmark-publisher", "sourceUrl": "https://arxiv.org/abs/2607.05462"}], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-06", "source": {"authors": ["Edwin H. Wintermute", "Harmon Bhasin", "Christina M. Agapakis", "Dianzhuo Wang", "Evan Seeyave", "Arjun Banerjee", "Daniel Fulop", "Matthew C. Watson", "Adam J. Meyer", "Sandrine Boissel", "Jens H. Kuhn", "Rishi Jain", "Noah D. Taylor", "Helena Shomar", "Patrick M. Boyle", "Kenny Workman"], "categories": ["cs.CR", "cs.AI"], "id": "2607.05462", "title": "BioSecBench-Refusal: A paired metric for performance and alignment in agentic biosecurity risk assessment", "type": "arxiv", "url": "https://arxiv.org/abs/2607.05462"}, "sourceUpdatedAt": "2026-07-21T17:05:39Z", "topics": ["Safety"], "venueAttempts": [], "whyItMatters": "Addresses the need for benchmarks that quantify both capability and safety in agentic biosecurity, helping developers calibrate models to avoid over-refusal while still detecting threats."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
