{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:P35KLMVJQGT4BBRORPTXFZS7KA","short_pith_number":"pith:P35KLMVJ","schema_version":"1.0","canonical_sha256":"7efaa5b2a981a7c0862e8be772e65f5001603d306b6c99c5e2f677eb1c521d27","source":{"kind":"arxiv","id":"2506.03723","version":1},"attestation_state":"computed","paper":{"title":"Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervision","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chaeyun Jang, Hyungi Lee, Juho Lee, Moonseok Choi, Yegon Kim","submitted_at":"2025-06-04T08:56:24Z","abstract_excerpt":"Uncertainty calibration is essential for the safe deployment of large language models (LLMs), particularly when users rely on verbalized confidence estimates. While prior work has focused on classifiers or short-form generation, confidence calibration for chain-of-thought (CoT) reasoning remains largely unexplored. Surprisingly, we find that supervised fine-tuning with scalar confidence labels alone suffices to elicit self-verification behavior of language models, without any explicit reasoning supervision or reinforcement learning-based rewards. Despite being trained only to produce a verbali"},"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":"2506.03723","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-04T08:56:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b1cbb16f2f1a29e7cc9a75a13b405deced3993ac052c812fbc2508da4964981c","abstract_canon_sha256":"f8d15fab0c1a39eedd0236be320395a3ae48394f5b5a76719ffeef1ebe6bd04b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:15:50.762065Z","signature_b64":"KeJ49awyeCS5My54ZhIUiz8FfJ58jXO3d5bhOfdn6MwbT33pPRTsP8v4jTJrj8e6taavaZPHCwPyq1Ee2hFkCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7efaa5b2a981a7c0862e8be772e65f5001603d306b6c99c5e2f677eb1c521d27","last_reissued_at":"2026-07-05T11:15:50.761565Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:15:50.761565Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervision","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chaeyun Jang, Hyungi Lee, Juho Lee, Moonseok Choi, Yegon Kim","submitted_at":"2025-06-04T08:56:24Z","abstract_excerpt":"Uncertainty calibration is essential for the safe deployment of large language models (LLMs), particularly when users rely on verbalized confidence estimates. While prior work has focused on classifiers or short-form generation, confidence calibration for chain-of-thought (CoT) reasoning remains largely unexplored. Surprisingly, we find that supervised fine-tuning with scalar confidence labels alone suffices to elicit self-verification behavior of language models, without any explicit reasoning supervision or reinforcement learning-based rewards. Despite being trained only to produce a verbali"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.03723","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/2506.03723/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":"2506.03723","created_at":"2026-07-05T11:15:50.761637+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.03723v1","created_at":"2026-07-05T11:15:50.761637+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.03723","created_at":"2026-07-05T11:15:50.761637+00:00"},{"alias_kind":"pith_short_12","alias_value":"P35KLMVJQGT4","created_at":"2026-07-05T11:15:50.761637+00:00"},{"alias_kind":"pith_short_16","alias_value":"P35KLMVJQGT4BBRO","created_at":"2026-07-05T11:15:50.761637+00:00"},{"alias_kind":"pith_short_8","alias_value":"P35KLMVJ","created_at":"2026-07-05T11:15:50.761637+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18089","citing_title":"From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning","ref_index":113,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03969","citing_title":"Quantifying Faithful Confidence Expression in Large Reasoning Models","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA","json":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA.json","graph_json":"https://pith.science/api/pith-number/P35KLMVJQGT4BBRORPTXFZS7KA/graph.json","events_json":"https://pith.science/api/pith-number/P35KLMVJQGT4BBRORPTXFZS7KA/events.json","paper":"https://pith.science/paper/P35KLMVJ"},"agent_actions":{"view_html":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA","download_json":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA.json","view_paper":"https://pith.science/paper/P35KLMVJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.03723&json=true","fetch_graph":"https://pith.science/api/pith-number/P35KLMVJQGT4BBRORPTXFZS7KA/graph.json","fetch_events":"https://pith.science/api/pith-number/P35KLMVJQGT4BBRORPTXFZS7KA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA/action/storage_attestation","attest_author":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA/action/author_attestation","sign_citation":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA/action/citation_signature","submit_replication":"https://pith.science/pith/P35KLMVJQGT4BBRORPTXFZS7KA/action/replication_record"}},"created_at":"2026-07-05T11:15:50.761637+00:00","updated_at":"2026-07-05T11:15:50.761637+00:00"}