{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ROICRIYVT7Z27WGRKK73O7YJI2","short_pith_number":"pith:ROICRIYV","schema_version":"1.0","canonical_sha256":"8b9028a3159ff3afd8d152bfb77f0946b345f1262c1ac3024ec7b5adbd4a2bd7","source":{"kind":"arxiv","id":"2412.16974","version":1},"attestation_state":"computed","paper":{"title":"Cannot or Should Not? Automatic Analysis of Refusal Composition in IFT/RLHF Datasets and Refusal Behavior of Black-Box LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Alexander von Recum, Christoph Schnabl, Gabor Hollbeck, Marvin Von Hagen, Philip Blinde, Silas Alberti","submitted_at":"2024-12-22T11:16:53Z","abstract_excerpt":"Refusals - instances where large language models (LLMs) decline or fail to fully execute user instructions - are crucial for both AI safety and AI capabilities and the reduction of hallucinations in particular. These behaviors are learned during post-training, especially in instruction fine-tuning (IFT) and reinforcement learning from human feedback (RLHF). However, existing taxonomies and evaluation datasets for refusals are inadequate, often focusing solely on should-not-related (instead of cannot-related) categories, and lacking tools for auditing refusal content in black-box LLM outputs.\n "},"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":"2412.16974","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-22T11:16:53Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"4aa2cdddca48481f3b02691cd36de3b853b31c803f01296e95aa90363627975e","abstract_canon_sha256":"ea628424a5c0183e1ed2e8a840290997eeec92ab831febb8c473eb9e604e029e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:53:15.772149Z","signature_b64":"VAadf3Gwl+Ea6v+o8ie4k0+9N1zlprGM4Cy2HEygt9Yr3CI+FfGiMbcVy+T0unRWk0q0j0v0SxZnz6HCCVG9CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8b9028a3159ff3afd8d152bfb77f0946b345f1262c1ac3024ec7b5adbd4a2bd7","last_reissued_at":"2026-07-05T09:53:15.771653Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:53:15.771653Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cannot or Should Not? Automatic Analysis of Refusal Composition in IFT/RLHF Datasets and Refusal Behavior of Black-Box LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Alexander von Recum, Christoph Schnabl, Gabor Hollbeck, Marvin Von Hagen, Philip Blinde, Silas Alberti","submitted_at":"2024-12-22T11:16:53Z","abstract_excerpt":"Refusals - instances where large language models (LLMs) decline or fail to fully execute user instructions - are crucial for both AI safety and AI capabilities and the reduction of hallucinations in particular. These behaviors are learned during post-training, especially in instruction fine-tuning (IFT) and reinforcement learning from human feedback (RLHF). However, existing taxonomies and evaluation datasets for refusals are inadequate, often focusing solely on should-not-related (instead of cannot-related) categories, and lacking tools for auditing refusal content in black-box LLM outputs.\n "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.16974","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/2412.16974/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":"2412.16974","created_at":"2026-07-05T09:53:15.771718+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.16974v1","created_at":"2026-07-05T09:53:15.771718+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.16974","created_at":"2026-07-05T09:53:15.771718+00:00"},{"alias_kind":"pith_short_12","alias_value":"ROICRIYVT7Z2","created_at":"2026-07-05T09:53:15.771718+00:00"},{"alias_kind":"pith_short_16","alias_value":"ROICRIYVT7Z27WGR","created_at":"2026-07-05T09:53:15.771718+00:00"},{"alias_kind":"pith_short_8","alias_value":"ROICRIYV","created_at":"2026-07-05T09:53:15.771718+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"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":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06233","citing_title":"Blind Refusal: Language Models Refuse to Help Users Evade Unjust, Absurd, and Illegitimate Rules","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2","json":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2.json","graph_json":"https://pith.science/api/pith-number/ROICRIYVT7Z27WGRKK73O7YJI2/graph.json","events_json":"https://pith.science/api/pith-number/ROICRIYVT7Z27WGRKK73O7YJI2/events.json","paper":"https://pith.science/paper/ROICRIYV"},"agent_actions":{"view_html":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2","download_json":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2.json","view_paper":"https://pith.science/paper/ROICRIYV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.16974&json=true","fetch_graph":"https://pith.science/api/pith-number/ROICRIYVT7Z27WGRKK73O7YJI2/graph.json","fetch_events":"https://pith.science/api/pith-number/ROICRIYVT7Z27WGRKK73O7YJI2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2/action/storage_attestation","attest_author":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2/action/author_attestation","sign_citation":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2/action/citation_signature","submit_replication":"https://pith.science/pith/ROICRIYVT7Z27WGRKK73O7YJI2/action/replication_record"}},"created_at":"2026-07-05T09:53:15.771718+00:00","updated_at":"2026-07-05T09:53:15.771718+00:00"}