{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZYKTUYB5LF6EU5KZUSMG77BWR4","short_pith_number":"pith:ZYKTUYB5","schema_version":"1.0","canonical_sha256":"ce153a603d597c4a7559a4986ffc368f2fd10f676f13dff899ea4fc2134b918f","source":{"kind":"arxiv","id":"2603.25112","version":2},"attestation_state":"computed","paper":{"title":"Do LLMs Know What They Know? Measuring Metacognitive Efficiency with Signal Detection Theory","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jon-Paul Cacioli","submitted_at":"2026-03-26T07:38:28Z","abstract_excerpt":"Standard evaluation of LLM confidence relies on calibration metrics (ECE, Brier score) that conflate how much a model knows (Type-1 accuracy) with how well its confidence signal tracks that knowledge (Type-2 metacognitive sensitivity). We apply Signal Detection Theory (SDT) to decompose these capacities, treating token-level normalised log-probability as a graded confidence variable and answer correctness as the state to be discriminated. We characterise the Type-2 ROC of this signal, including its unequal-variance structure via z-ROC analysis, and -- because the meta-d' efficiency ratio is no"},"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":"2603.25112","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T07:38:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1c5a184399c064dea72fd8d4ee2827a261bbf840f3aecc2ebf12b05da71eae91","abstract_canon_sha256":"240b9b134922941c225ffd12e555539ce0c9b4d757052ea9075022034ec974f6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-15T00:21:18.116289Z","signature_b64":"dlVoMA1akjSaJ+lT9r/8ntNmtwwws2IpXyKZReq5drK5yNFa5CN0oWqRmTH66LAeYDRDQ5pNLN6lbkjtkrKCDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ce153a603d597c4a7559a4986ffc368f2fd10f676f13dff899ea4fc2134b918f","last_reissued_at":"2026-07-15T00:21:18.115410Z","signature_status":"signed_v1","first_computed_at":"2026-07-15T00:21:18.115410Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do LLMs Know What They Know? Measuring Metacognitive Efficiency with Signal Detection Theory","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jon-Paul Cacioli","submitted_at":"2026-03-26T07:38:28Z","abstract_excerpt":"Standard evaluation of LLM confidence relies on calibration metrics (ECE, Brier score) that conflate how much a model knows (Type-1 accuracy) with how well its confidence signal tracks that knowledge (Type-2 metacognitive sensitivity). We apply Signal Detection Theory (SDT) to decompose these capacities, treating token-level normalised log-probability as a graded confidence variable and answer correctness as the state to be discriminated. We characterise the Type-2 ROC of this signal, including its unequal-variance structure via z-ROC analysis, and -- because the meta-d' efficiency ratio is no"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.25112","kind":"arxiv","version":2},"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/2603.25112/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":"2603.25112","created_at":"2026-07-15T00:21:18.115858+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.25112v2","created_at":"2026-07-15T00:21:18.115858+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.25112","created_at":"2026-07-15T00:21:18.115858+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZYKTUYB5LF6E","created_at":"2026-07-15T00:21:18.115858+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZYKTUYB5LF6EU5KZ","created_at":"2026-07-15T00:21:18.115858+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZYKTUYB5","created_at":"2026-07-15T00:21:18.115858+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":7,"sample":[{"citing_arxiv_id":"2605.14859","citing_title":"Do Coding Agents Understand Least-Privilege Authorization?","ref_index":71,"is_internal_anchor":true},{"citing_arxiv_id":"2604.22215","citing_title":"Verbal Confidence Saturation in 3-9B Open-Weight Instruction-Tuned LLMs: A Pre-Registered Psychometric Validity Screen","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2604.11011","citing_title":"K-Way Energy Probes for Metacognition Reduce to Softmax in Discriminative Predictive Coding Networks","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08976","citing_title":"Quantisation Reshapes the Metacognitive Geometry of Language Models","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2604.16009","citing_title":"MEDLEY-BENCH: Scale Buys Evaluation but Not Control in AI Metacognition","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2604.17707","citing_title":"Before You Interpret the Profile: Validity Scaling for LLM Metacognitive Self-Report","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05851","citing_title":"Hypothesis generation and updating in large language models","ref_index":27,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4","json":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4.json","graph_json":"https://pith.science/api/pith-number/ZYKTUYB5LF6EU5KZUSMG77BWR4/graph.json","events_json":"https://pith.science/api/pith-number/ZYKTUYB5LF6EU5KZUSMG77BWR4/events.json","paper":"https://pith.science/paper/ZYKTUYB5"},"agent_actions":{"view_html":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4","download_json":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4.json","view_paper":"https://pith.science/paper/ZYKTUYB5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.25112&json=true","fetch_graph":"https://pith.science/api/pith-number/ZYKTUYB5LF6EU5KZUSMG77BWR4/graph.json","fetch_events":"https://pith.science/api/pith-number/ZYKTUYB5LF6EU5KZUSMG77BWR4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4/action/storage_attestation","attest_author":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4/action/author_attestation","sign_citation":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4/action/citation_signature","submit_replication":"https://pith.science/pith/ZYKTUYB5LF6EU5KZUSMG77BWR4/action/replication_record"}},"created_at":"2026-07-15T00:21:18.115858+00:00","updated_at":"2026-07-15T00:21:18.115858+00:00"}