{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5D2NVMPKVCIL425HRUZS6WP6QL","short_pith_number":"pith:5D2NVMPK","schema_version":"1.0","canonical_sha256":"e8f4dab1eaa890be6ba78d332f59fe82e1930c1ca0117b7f2b0f0380db0bc7f0","source":{"kind":"arxiv","id":"2603.22161","version":2},"attestation_state":"computed","paper":{"title":"Causal Evidence that Language Models use Confidence to Drive Behavior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dharshan Kumaran, Nathaniel Daw, Petar Veli\\v{c}kovi\\'c, Simon Osindero, Viorica Patraucean","submitted_at":"2026-03-23T16:23:31Z","abstract_excerpt":"Metacognition -- assessing the quality of one's own cognitive performance -- guides adaptive behavior across species. Substantial research demonstrates that confidence signals can be extracted from language model outputs, yet a fundamental question remains: do models actually use these signals to control behavior, such as deciding whether to answer or abstain? To investigate, we developed a four-phase paradigm. Phase~1 elicited baseline confidence estimates without an abstention option. Phase~2 revealed that LLMs apply an implicit threshold to internal confidence when deciding to abstain, with"},"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":"2603.22161","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-23T16:23:31Z","cross_cats_sorted":[],"title_canon_sha256":"30fc68310199aacb7f9d954a081f62edf10a8c693e9f7574de254a7e59e975ac","abstract_canon_sha256":"3dfbe25958b14084619a27b6272d140a40b6199ff75e43da83fde4e63e642c80"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:42.311232Z","signature_b64":"vGcMogEgEQ6jbiyLFd/Cj4xaZdCJXA1YSjdBHUzQQeXi187gA8CYgMT1iRjHC+H3SEwC2Xb2D7ui5jMpqoXCDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e8f4dab1eaa890be6ba78d332f59fe82e1930c1ca0117b7f2b0f0380db0bc7f0","last_reissued_at":"2026-05-20T02:05:42.310248Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:42.310248Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Causal Evidence that Language Models use Confidence to Drive Behavior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dharshan Kumaran, Nathaniel Daw, Petar Veli\\v{c}kovi\\'c, Simon Osindero, Viorica Patraucean","submitted_at":"2026-03-23T16:23:31Z","abstract_excerpt":"Metacognition -- assessing the quality of one's own cognitive performance -- guides adaptive behavior across species. Substantial research demonstrates that confidence signals can be extracted from language model outputs, yet a fundamental question remains: do models actually use these signals to control behavior, such as deciding whether to answer or abstain? To investigate, we developed a four-phase paradigm. Phase~1 elicited baseline confidence estimates without an abstention option. Phase~2 revealed that LLMs apply an implicit threshold to internal confidence when deciding to abstain, with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.22161","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/2603.22161/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":"2603.22161","created_at":"2026-05-20T02:05:42.310426+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.22161v2","created_at":"2026-05-20T02:05:42.310426+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.22161","created_at":"2026-05-20T02:05:42.310426+00:00"},{"alias_kind":"pith_short_12","alias_value":"5D2NVMPKVCIL","created_at":"2026-05-20T02:05:42.310426+00:00"},{"alias_kind":"pith_short_16","alias_value":"5D2NVMPKVCIL425H","created_at":"2026-05-20T02:05:42.310426+00:00"},{"alias_kind":"pith_short_8","alias_value":"5D2NVMPK","created_at":"2026-05-20T02:05:42.310426+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/5D2NVMPKVCIL425HRUZS6WP6QL","json":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL.json","graph_json":"https://pith.science/api/pith-number/5D2NVMPKVCIL425HRUZS6WP6QL/graph.json","events_json":"https://pith.science/api/pith-number/5D2NVMPKVCIL425HRUZS6WP6QL/events.json","paper":"https://pith.science/paper/5D2NVMPK"},"agent_actions":{"view_html":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL","download_json":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL.json","view_paper":"https://pith.science/paper/5D2NVMPK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.22161&json=true","fetch_graph":"https://pith.science/api/pith-number/5D2NVMPKVCIL425HRUZS6WP6QL/graph.json","fetch_events":"https://pith.science/api/pith-number/5D2NVMPKVCIL425HRUZS6WP6QL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL/action/storage_attestation","attest_author":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL/action/author_attestation","sign_citation":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL/action/citation_signature","submit_replication":"https://pith.science/pith/5D2NVMPKVCIL425HRUZS6WP6QL/action/replication_record"}},"created_at":"2026-05-20T02:05:42.310426+00:00","updated_at":"2026-05-20T02:05:42.310426+00:00"}