{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2MGVWQN33P5MWN4Y4HGIRY5DJL","short_pith_number":"pith:2MGVWQN3","schema_version":"1.0","canonical_sha256":"d30d5b41bbdbfacb3798e1cc88e3a34ac3a3e49e94f48b7a832d5e459a511e79","source":{"kind":"arxiv","id":"2507.20252","version":3},"attestation_state":"computed","paper":{"title":"Post-Completion Learning for Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Can Huang, Chao Feng, Hao Feng, Shu Wei, Siqi Wang, Wei Shi, Xiang Fei, Yuxiang Nie","submitted_at":"2025-07-27T12:47:26Z","abstract_excerpt":"Current language model training paradigms typically terminate learning upon reaching the end-of-sequence (<eos>) token, overlooking the potential learning opportunities in the post-completion space. We propose Post-Completion Learning (PCL), a novel training framework that systematically utilizes the sequence space after model output completion, to enhance both the reasoning and self-evaluation abilities. PCL enables models to continue generating self-assessments and reward predictions during training, while maintaining efficient inference by stopping at the completion point.\n  To fully utiliz"},"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":"2507.20252","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-27T12:47:26Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"67377577acbb6af3eba103993edf3252d7935b0cfc340b91959fe9e2060a4dc8","abstract_canon_sha256":"bf21523ca5ed57c97873d31153cc585e0260af2b98dcdc2bc36a36cde6799965"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:52:22.361703Z","signature_b64":"IfCTckB7YOGP5DmxSZw3BEN7C39Fnf8dXDgx2YESMJhbzvZ60rZ60t6zPSLsC8AYjEpjYNJqUyYXtWmGrMUNBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d30d5b41bbdbfacb3798e1cc88e3a34ac3a3e49e94f48b7a832d5e459a511e79","last_reissued_at":"2026-07-05T11:52:22.361205Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:52:22.361205Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Post-Completion Learning for Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Can Huang, Chao Feng, Hao Feng, Shu Wei, Siqi Wang, Wei Shi, Xiang Fei, Yuxiang Nie","submitted_at":"2025-07-27T12:47:26Z","abstract_excerpt":"Current language model training paradigms typically terminate learning upon reaching the end-of-sequence (<eos>) token, overlooking the potential learning opportunities in the post-completion space. We propose Post-Completion Learning (PCL), a novel training framework that systematically utilizes the sequence space after model output completion, to enhance both the reasoning and self-evaluation abilities. PCL enables models to continue generating self-assessments and reward predictions during training, while maintaining efficient inference by stopping at the completion point.\n  To fully utiliz"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.20252","kind":"arxiv","version":3},"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/2507.20252/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":"2507.20252","created_at":"2026-07-05T11:52:22.361264+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.20252v3","created_at":"2026-07-05T11:52:22.361264+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.20252","created_at":"2026-07-05T11:52:22.361264+00:00"},{"alias_kind":"pith_short_12","alias_value":"2MGVWQN33P5M","created_at":"2026-07-05T11:52:22.361264+00:00"},{"alias_kind":"pith_short_16","alias_value":"2MGVWQN33P5MWN4Y","created_at":"2026-07-05T11:52:22.361264+00:00"},{"alias_kind":"pith_short_8","alias_value":"2MGVWQN3","created_at":"2026-07-05T11:52:22.361264+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01249","citing_title":"Trust Region On-Policy Distillation","ref_index":174,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":139,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL","json":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL.json","graph_json":"https://pith.science/api/pith-number/2MGVWQN33P5MWN4Y4HGIRY5DJL/graph.json","events_json":"https://pith.science/api/pith-number/2MGVWQN33P5MWN4Y4HGIRY5DJL/events.json","paper":"https://pith.science/paper/2MGVWQN3"},"agent_actions":{"view_html":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL","download_json":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL.json","view_paper":"https://pith.science/paper/2MGVWQN3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.20252&json=true","fetch_graph":"https://pith.science/api/pith-number/2MGVWQN33P5MWN4Y4HGIRY5DJL/graph.json","fetch_events":"https://pith.science/api/pith-number/2MGVWQN33P5MWN4Y4HGIRY5DJL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL/action/storage_attestation","attest_author":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL/action/author_attestation","sign_citation":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL/action/citation_signature","submit_replication":"https://pith.science/pith/2MGVWQN33P5MWN4Y4HGIRY5DJL/action/replication_record"}},"created_at":"2026-07-05T11:52:22.361264+00:00","updated_at":"2026-07-05T11:52:22.361264+00:00"}