{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ENJY3RVY4RIP5ABA4ZEEFT7K7O","short_pith_number":"pith:ENJY3RVY","schema_version":"1.0","canonical_sha256":"23538dc6b8e450fe8020e64842cfeafb99c919524a1a8d234c2640366435b49f","source":{"kind":"arxiv","id":"2405.18634","version":2},"attestation_state":"computed","paper":{"title":"A Theoretical Understanding of Self-Correction through In-context Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Stefanie Jegelka, Yifei Wang, Yisen Wang, Yuyang Wu, Zeming Wei","submitted_at":"2024-05-28T22:33:02Z","abstract_excerpt":"Going beyond mimicking limited human experiences, recent studies show initial evidence that, like humans, large language models (LLMs) are capable of improving their abilities purely by self-correction, i.e., correcting previous responses through self-examination, in certain circumstances. Nevertheless, little is known about how such capabilities arise. In this work, based on a simplified setup akin to an alignment task, we theoretically analyze self-correction from an in-context learning perspective, showing that when LLMs give relatively accurate self-examinations as rewards, they are capabl"},"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":"2405.18634","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-28T22:33:02Z","cross_cats_sorted":["cs.CL","stat.ML"],"title_canon_sha256":"5de167b9fb4f7e2a7f23aea3645c72db50413b613059e2bedf2fde65f23a950c","abstract_canon_sha256":"5afa88311632f4d2d90db84b4bf5d0ef977cba810d50817ef555fe7e4782ef6a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:36:41.195240Z","signature_b64":"wrzBzz3cuJ4L8cKyNZQ30FEe+AyXDyepfPQzjtqHPdJORF0So5XSJ1zmWLhSZeoeBWIiOokv0NOHJLaJ+3HuCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"23538dc6b8e450fe8020e64842cfeafb99c919524a1a8d234c2640366435b49f","last_reissued_at":"2026-07-05T09:36:41.194824Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:36:41.194824Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Theoretical Understanding of Self-Correction through In-context Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Stefanie Jegelka, Yifei Wang, Yisen Wang, Yuyang Wu, Zeming Wei","submitted_at":"2024-05-28T22:33:02Z","abstract_excerpt":"Going beyond mimicking limited human experiences, recent studies show initial evidence that, like humans, large language models (LLMs) are capable of improving their abilities purely by self-correction, i.e., correcting previous responses through self-examination, in certain circumstances. Nevertheless, little is known about how such capabilities arise. In this work, based on a simplified setup akin to an alignment task, we theoretically analyze self-correction from an in-context learning perspective, showing that when LLMs give relatively accurate self-examinations as rewards, they are capabl"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18634","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/2405.18634/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":"2405.18634","created_at":"2026-07-05T09:36:41.194882+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18634v2","created_at":"2026-07-05T09:36:41.194882+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18634","created_at":"2026-07-05T09:36:41.194882+00:00"},{"alias_kind":"pith_short_12","alias_value":"ENJY3RVY4RIP","created_at":"2026-07-05T09:36:41.194882+00:00"},{"alias_kind":"pith_short_16","alias_value":"ENJY3RVY4RIP5ABA","created_at":"2026-07-05T09:36:41.194882+00:00"},{"alias_kind":"pith_short_8","alias_value":"ENJY3RVY","created_at":"2026-07-05T09:36:41.194882+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.17433","citing_title":"Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O","json":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O.json","graph_json":"https://pith.science/api/pith-number/ENJY3RVY4RIP5ABA4ZEEFT7K7O/graph.json","events_json":"https://pith.science/api/pith-number/ENJY3RVY4RIP5ABA4ZEEFT7K7O/events.json","paper":"https://pith.science/paper/ENJY3RVY"},"agent_actions":{"view_html":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O","download_json":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O.json","view_paper":"https://pith.science/paper/ENJY3RVY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18634&json=true","fetch_graph":"https://pith.science/api/pith-number/ENJY3RVY4RIP5ABA4ZEEFT7K7O/graph.json","fetch_events":"https://pith.science/api/pith-number/ENJY3RVY4RIP5ABA4ZEEFT7K7O/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O/action/storage_attestation","attest_author":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O/action/author_attestation","sign_citation":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O/action/citation_signature","submit_replication":"https://pith.science/pith/ENJY3RVY4RIP5ABA4ZEEFT7K7O/action/replication_record"}},"created_at":"2026-07-05T09:36:41.194882+00:00","updated_at":"2026-07-05T09:36:41.194882+00:00"}