{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LIUNAH6NJHXJHR4NH6VX7VMATS","short_pith_number":"pith:LIUNAH6N","schema_version":"1.0","canonical_sha256":"5a28d01fcd49ee93c78d3fab7fd5809c8624676faffa3494d938a222fdb4c3a6","source":{"kind":"arxiv","id":"2502.06867","version":1},"attestation_state":"computed","paper":{"title":"Forbidden Science: Dual-Use AI Challenge Benchmark and Scientific Refusal Tests","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Noever, Forrest McKee","submitted_at":"2025-02-08T04:27:33Z","abstract_excerpt":"The development of robust safety benchmarks for large language models requires open, reproducible datasets that can measure both appropriate refusal of harmful content and potential over-restriction of legitimate scientific discourse. We present an open-source dataset and testing framework for evaluating LLM safety mechanisms across mainly controlled substance queries, analyzing four major models' responses to systematically varied prompts. Our results reveal distinct safety profiles: Claude-3.5-sonnet demonstrated the most conservative approach with 73% refusals and 27% allowances, while Mist"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.06867","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-08T04:27:33Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"4198290f1b151038215c123bb4722c9e610ed3cc2eedbb1b5267fc6668a121d7","abstract_canon_sha256":"303fce607e2bec3a806c3d9068e5ba597302612ae76e5d04e2015beae03d3dac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:12:29.109026Z","signature_b64":"mjjtzb8WUK9AW0TWmfs6Yl3wklHLwRRNrftuo1Z1WCmxz5zGk62lTLuU19OIIjpICcehsQua/VLvUffgAxziAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5a28d01fcd49ee93c78d3fab7fd5809c8624676faffa3494d938a222fdb4c3a6","last_reissued_at":"2026-07-05T10:12:29.108559Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:12:29.108559Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Forbidden Science: Dual-Use AI Challenge Benchmark and Scientific Refusal Tests","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Noever, Forrest McKee","submitted_at":"2025-02-08T04:27:33Z","abstract_excerpt":"The development of robust safety benchmarks for large language models requires open, reproducible datasets that can measure both appropriate refusal of harmful content and potential over-restriction of legitimate scientific discourse. We present an open-source dataset and testing framework for evaluating LLM safety mechanisms across mainly controlled substance queries, analyzing four major models' responses to systematically varied prompts. Our results reveal distinct safety profiles: Claude-3.5-sonnet demonstrated the most conservative approach with 73% refusals and 27% allowances, while Mist"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.06867","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2502.06867/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.06867","created_at":"2026-07-05T10:12:29.108616+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.06867v1","created_at":"2026-07-05T10:12:29.108616+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.06867","created_at":"2026-07-05T10:12:29.108616+00:00"},{"alias_kind":"pith_short_12","alias_value":"LIUNAH6NJHXJ","created_at":"2026-07-05T10:12:29.108616+00:00"},{"alias_kind":"pith_short_16","alias_value":"LIUNAH6NJHXJHR4N","created_at":"2026-07-05T10:12:29.108616+00:00"},{"alias_kind":"pith_short_8","alias_value":"LIUNAH6N","created_at":"2026-07-05T10:12:29.108616+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.21545","citing_title":"RefusalBench: Why Refusal Rate Misranks Frontier LLMs on Biological Research Prompts","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS","json":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS.json","graph_json":"https://pith.science/api/pith-number/LIUNAH6NJHXJHR4NH6VX7VMATS/graph.json","events_json":"https://pith.science/api/pith-number/LIUNAH6NJHXJHR4NH6VX7VMATS/events.json","paper":"https://pith.science/paper/LIUNAH6N"},"agent_actions":{"view_html":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS","download_json":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS.json","view_paper":"https://pith.science/paper/LIUNAH6N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.06867&json=true","fetch_graph":"https://pith.science/api/pith-number/LIUNAH6NJHXJHR4NH6VX7VMATS/graph.json","fetch_events":"https://pith.science/api/pith-number/LIUNAH6NJHXJHR4NH6VX7VMATS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS/action/storage_attestation","attest_author":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS/action/author_attestation","sign_citation":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS/action/citation_signature","submit_replication":"https://pith.science/pith/LIUNAH6NJHXJHR4NH6VX7VMATS/action/replication_record"}},"created_at":"2026-07-05T10:12:29.108616+00:00","updated_at":"2026-07-05T10:12:29.108616+00:00"}