{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:YFHEPNYGWDO7AD7LJTCLUC2BYW","short_pith_number":"pith:YFHEPNYG","schema_version":"1.0","canonical_sha256":"c14e47b706b0ddf00feb4cc4ba0b41c59659c379e1a543715a86bf6847a2d77f","source":{"kind":"arxiv","id":"2607.08456","version":1},"attestation_state":"computed","paper":{"title":"Two Axes of LLM Abstention: Answer Correctness and Question Answerability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Benedikt J. Wagner","submitted_at":"2026-07-09T13:17:56Z","abstract_excerpt":"A model should refuse two different things: answers it would get wrong, and questions it should not answer at all, such as unanswerable ones or ones resting on a false premise. The usual recipe thresholds a single confidence score, which cannot tell these apart. Across five instruction-tuned models from three families (2B to 14B), we find they are separate axes. Ordinary answer-confidence tracks whether an answer is right but is nearly blind to whether the question is answerable; a linear probe on hidden states does the reverse. The blind spot does not shrink with scale. It is worst on natural"},"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":"2607.08456","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-09T13:17:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c1f3d26a70084956ebd30491fbbf5a99822e47e9b319bf667ab2dc0751849e5f","abstract_canon_sha256":"58cc0b1f01fde30805b1187797dff5e5a7d8dee487083febd9cdf21c8a5d245a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-10T01:19:51.140013Z","signature_b64":"nFPLTw1jVajAgPJ0zGAZLjShaGK51s//haL66GiKD4G+Q4OZGTfQV7c/+kZ1h3eK7HfRo1AxaKH5asHrgU96Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c14e47b706b0ddf00feb4cc4ba0b41c59659c379e1a543715a86bf6847a2d77f","last_reissued_at":"2026-07-10T01:19:51.139606Z","signature_status":"signed_v1","first_computed_at":"2026-07-10T01:19:51.139606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Two Axes of LLM Abstention: Answer Correctness and Question Answerability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Benedikt J. Wagner","submitted_at":"2026-07-09T13:17:56Z","abstract_excerpt":"A model should refuse two different things: answers it would get wrong, and questions it should not answer at all, such as unanswerable ones or ones resting on a false premise. The usual recipe thresholds a single confidence score, which cannot tell these apart. Across five instruction-tuned models from three families (2B to 14B), we find they are separate axes. Ordinary answer-confidence tracks whether an answer is right but is nearly blind to whether the question is answerable; a linear probe on hidden states does the reverse. The blind spot does not shrink with scale. It is worst on natural"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.08456","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/2607.08456/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":"2607.08456","created_at":"2026-07-10T01:19:51.139664+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.08456v1","created_at":"2026-07-10T01:19:51.139664+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.08456","created_at":"2026-07-10T01:19:51.139664+00:00"},{"alias_kind":"pith_short_12","alias_value":"YFHEPNYGWDO7","created_at":"2026-07-10T01:19:51.139664+00:00"},{"alias_kind":"pith_short_16","alias_value":"YFHEPNYGWDO7AD7L","created_at":"2026-07-10T01:19:51.139664+00:00"},{"alias_kind":"pith_short_8","alias_value":"YFHEPNYG","created_at":"2026-07-10T01:19:51.139664+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW","json":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW.json","graph_json":"https://pith.science/api/pith-number/YFHEPNYGWDO7AD7LJTCLUC2BYW/graph.json","events_json":"https://pith.science/api/pith-number/YFHEPNYGWDO7AD7LJTCLUC2BYW/events.json","paper":"https://pith.science/paper/YFHEPNYG"},"agent_actions":{"view_html":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW","download_json":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW.json","view_paper":"https://pith.science/paper/YFHEPNYG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.08456&json=true","fetch_graph":"https://pith.science/api/pith-number/YFHEPNYGWDO7AD7LJTCLUC2BYW/graph.json","fetch_events":"https://pith.science/api/pith-number/YFHEPNYGWDO7AD7LJTCLUC2BYW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW/action/storage_attestation","attest_author":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW/action/author_attestation","sign_citation":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW/action/citation_signature","submit_replication":"https://pith.science/pith/YFHEPNYGWDO7AD7LJTCLUC2BYW/action/replication_record"}},"created_at":"2026-07-10T01:19:51.139664+00:00","updated_at":"2026-07-10T01:19:51.139664+00:00"}