{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M243X7DS6KY7S2FU2JNCQVP5Y5","short_pith_number":"pith:M243X7DS","schema_version":"1.0","canonical_sha256":"66b9bbfc72f2b1f968b4d25a2855fdc75c1d774d21d7f483ca3214498127792f","source":{"kind":"arxiv","id":"2508.14460","version":1},"attestation_state":"computed","paper":{"title":"DuPO: Enabling Reliable LLM Self-Verification via Dual Preference Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Lu Lu, Lu Xu, Shanbo Cheng, Shuaijie She, Shujian Huang, Tao Li, Wenhao Zhu, Yu Bao, Yu Lu, Yuxuan Wang","submitted_at":"2025-08-20T06:31:18Z","abstract_excerpt":"We present DuPO, a dual learning-based preference optimization framework that generates annotation-free feedback via a generalized duality. DuPO addresses two key limitations: Reinforcement Learning with Verifiable Rewards (RLVR)'s reliance on costly labels and applicability restricted to verifiable tasks, and traditional dual learning's restriction to strictly dual task pairs (e.g., translation and back-translation). Specifically, DuPO decomposes a primal task's input into known and unknown components, then constructs its dual task to reconstruct the unknown part using the primal output and k"},"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":"2508.14460","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-08-20T06:31:18Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"78d4017885c0bdee48f3eac76ef440f3eb2657c2197805c7fa1f014bc40a8bfa","abstract_canon_sha256":"6444a1b1dac105cb0ce44c8251662f13e3c698582b6ad8d67b8c502cd8c33adc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:56:32.411784Z","signature_b64":"UG/imMw6OtL7CU0t4h0taYEFl5AvsFZTxYS37PnIIwmOlNAB+rF900f21Ws+SP/IeBgVZ3+Eh9S8CZraWUj0DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"66b9bbfc72f2b1f968b4d25a2855fdc75c1d774d21d7f483ca3214498127792f","last_reissued_at":"2026-07-05T11:56:32.411309Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:56:32.411309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DuPO: Enabling Reliable LLM Self-Verification via Dual Preference Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Lu Lu, Lu Xu, Shanbo Cheng, Shuaijie She, Shujian Huang, Tao Li, Wenhao Zhu, Yu Bao, Yu Lu, Yuxuan Wang","submitted_at":"2025-08-20T06:31:18Z","abstract_excerpt":"We present DuPO, a dual learning-based preference optimization framework that generates annotation-free feedback via a generalized duality. DuPO addresses two key limitations: Reinforcement Learning with Verifiable Rewards (RLVR)'s reliance on costly labels and applicability restricted to verifiable tasks, and traditional dual learning's restriction to strictly dual task pairs (e.g., translation and back-translation). Specifically, DuPO decomposes a primal task's input into known and unknown components, then constructs its dual task to reconstruct the unknown part using the primal output and k"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.14460","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/2508.14460/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":"2508.14460","created_at":"2026-07-05T11:56:32.411381+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.14460v1","created_at":"2026-07-05T11:56:32.411381+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.14460","created_at":"2026-07-05T11:56:32.411381+00:00"},{"alias_kind":"pith_short_12","alias_value":"M243X7DS6KY7","created_at":"2026-07-05T11:56:32.411381+00:00"},{"alias_kind":"pith_short_16","alias_value":"M243X7DS6KY7S2FU","created_at":"2026-07-05T11:56:32.411381+00:00"},{"alias_kind":"pith_short_8","alias_value":"M243X7DS","created_at":"2026-07-05T11:56:32.411381+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07663","citing_title":"Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops","ref_index":147,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18961","citing_title":"Be Your Own Teacher: Steering Protein Language Models via Unsupervised Reward Optimization","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2509.02547","citing_title":"The Landscape of Agentic Reinforcement Learning for LLMs: A Survey","ref_index":168,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12943","citing_title":"Reinforced Collaboration in Multi-Agent Flow Networks","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17928","citing_title":"HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5","json":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5.json","graph_json":"https://pith.science/api/pith-number/M243X7DS6KY7S2FU2JNCQVP5Y5/graph.json","events_json":"https://pith.science/api/pith-number/M243X7DS6KY7S2FU2JNCQVP5Y5/events.json","paper":"https://pith.science/paper/M243X7DS"},"agent_actions":{"view_html":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5","download_json":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5.json","view_paper":"https://pith.science/paper/M243X7DS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.14460&json=true","fetch_graph":"https://pith.science/api/pith-number/M243X7DS6KY7S2FU2JNCQVP5Y5/graph.json","fetch_events":"https://pith.science/api/pith-number/M243X7DS6KY7S2FU2JNCQVP5Y5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5/action/storage_attestation","attest_author":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5/action/author_attestation","sign_citation":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5/action/citation_signature","submit_replication":"https://pith.science/pith/M243X7DS6KY7S2FU2JNCQVP5Y5/action/replication_record"}},"created_at":"2026-07-05T11:56:32.411381+00:00","updated_at":"2026-07-05T11:56:32.411381+00:00"}