{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XNCMYB5T7G72OQYTVJXA3U6ABX","short_pith_number":"pith:XNCMYB5T","schema_version":"1.0","canonical_sha256":"bb44cc07b3f9bfa74313aa6e0dd3c00dc2894711b6c55939de39b03e68599ea9","source":{"kind":"arxiv","id":"2606.18327","version":1},"attestation_state":"computed","paper":{"title":"Self-CTRL: Self-Consistency Training with Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Belinda Z. Li, Itamar Pres, Jacob Andreas, Laura Ruis, Melat Ghebreselassie","submitted_at":"2026-06-16T17:59:40Z","abstract_excerpt":"Language models (LMs) that faithfully describe their own behavior can more easily be audited, understood, and trusted by users. This paper describes Self-Consistency Training with Reinforcement Learning (Self-CTRL), a method that optimizes for consistency between a LM's self-explanations and behavior on related inputs by updating explanations to better predict behavior or updating behavior to better match explanations. We apply our method in two domains. First, we study a formal probabilistic reasoning task in which LMs must learn to imitate a family of biased samplers and evaluated on their a"},"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":"2606.18327","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-16T17:59:40Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b17d5890c8e6ac06ff1bb33a539f36822685a6ad21ab9a637054e13e028cfa26","abstract_canon_sha256":"910c3e09921237a83904ad0dd3f1a0c7b5f978aef491857b5c8bd368b255407e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:10:57.987885Z","signature_b64":"98nitoEyv5zdiwmr81jI3V/OCcsB2pxxMbv1h11wr4JWyohIV2CacLcbPOnM7d7kFVDEg0dgdnvSWH3lZFUFAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bb44cc07b3f9bfa74313aa6e0dd3c00dc2894711b6c55939de39b03e68599ea9","last_reissued_at":"2026-06-19T16:10:57.987518Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:10:57.987518Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-CTRL: Self-Consistency Training with Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Belinda Z. Li, Itamar Pres, Jacob Andreas, Laura Ruis, Melat Ghebreselassie","submitted_at":"2026-06-16T17:59:40Z","abstract_excerpt":"Language models (LMs) that faithfully describe their own behavior can more easily be audited, understood, and trusted by users. This paper describes Self-Consistency Training with Reinforcement Learning (Self-CTRL), a method that optimizes for consistency between a LM's self-explanations and behavior on related inputs by updating explanations to better predict behavior or updating behavior to better match explanations. We apply our method in two domains. First, we study a formal probabilistic reasoning task in which LMs must learn to imitate a family of biased samplers and evaluated on their a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.18327","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/2606.18327/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":"2606.18327","created_at":"2026-06-19T16:10:57.987580+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.18327v1","created_at":"2026-06-19T16:10:57.987580+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.18327","created_at":"2026-06-19T16:10:57.987580+00:00"},{"alias_kind":"pith_short_12","alias_value":"XNCMYB5T7G72","created_at":"2026-06-19T16:10:57.987580+00:00"},{"alias_kind":"pith_short_16","alias_value":"XNCMYB5T7G72OQYT","created_at":"2026-06-19T16:10:57.987580+00:00"},{"alias_kind":"pith_short_8","alias_value":"XNCMYB5T","created_at":"2026-06-19T16:10:57.987580+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/XNCMYB5T7G72OQYTVJXA3U6ABX","json":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX.json","graph_json":"https://pith.science/api/pith-number/XNCMYB5T7G72OQYTVJXA3U6ABX/graph.json","events_json":"https://pith.science/api/pith-number/XNCMYB5T7G72OQYTVJXA3U6ABX/events.json","paper":"https://pith.science/paper/XNCMYB5T"},"agent_actions":{"view_html":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX","download_json":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX.json","view_paper":"https://pith.science/paper/XNCMYB5T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.18327&json=true","fetch_graph":"https://pith.science/api/pith-number/XNCMYB5T7G72OQYTVJXA3U6ABX/graph.json","fetch_events":"https://pith.science/api/pith-number/XNCMYB5T7G72OQYTVJXA3U6ABX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX/action/storage_attestation","attest_author":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX/action/author_attestation","sign_citation":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX/action/citation_signature","submit_replication":"https://pith.science/pith/XNCMYB5T7G72OQYTVJXA3U6ABX/action/replication_record"}},"created_at":"2026-06-19T16:10:57.987580+00:00","updated_at":"2026-06-19T16:10:57.987580+00:00"}