{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5UTQ3P6RDSIBFKATPGBSEW4ZAX","short_pith_number":"pith:5UTQ3P6R","schema_version":"1.0","canonical_sha256":"ed270dbfd11c9012a8137983225b9905fa0016fbefabe0a0333d06e583e6b988","source":{"kind":"arxiv","id":"2503.12854","version":3},"attestation_state":"computed","paper":{"title":"Enhancing LLM Reasoning with Iterative DPO: A Comprehensive Empirical Investigation","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongbin Zhao, Dongmei Jiang, Jiahao Lin, Linjing Li, Nan Xu, Qichao Zhang, Songjun Tu, Wei He, Xiangyuan Lan, Xiangyu Tian, Yuqian Fu","submitted_at":"2025-03-17T06:28:25Z","abstract_excerpt":"Recent advancements in post-training methodologies for large language models (LLMs) have highlighted reinforcement learning (RL) as a critical component for enhancing reasoning. However, the substantial computational costs associated with RL-based approaches have led to growing interest in alternative paradigms, such as Direct Preference Optimization (DPO). In this study, we investigate the effectiveness of DPO in facilitating self-improvement for LLMs through iterative preference-based learning. We demonstrate that a single round of DPO with coarse filtering significantly enhances mathematica"},"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":"2503.12854","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-17T06:28:25Z","cross_cats_sorted":[],"title_canon_sha256":"e0010db9d6ac1fd95fdb743da6ba443455b84fdb64b65e1bfc143db76e44b2ab","abstract_canon_sha256":"21754a4c1746339c351ea496ecb4892af7191d51907581c4a3184223bc5d5145"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:44:02.790364Z","signature_b64":"QlHUpDPlXnw2aCiSNceHnre2GQcngLqpvfI4trltukebUHgfjVwLqdU3K4VlD1HV/73K4tW+KsXje6u+bPKqCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed270dbfd11c9012a8137983225b9905fa0016fbefabe0a0333d06e583e6b988","last_reissued_at":"2026-07-05T11:44:02.789860Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:44:02.789860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enhancing LLM Reasoning with Iterative DPO: A Comprehensive Empirical Investigation","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongbin Zhao, Dongmei Jiang, Jiahao Lin, Linjing Li, Nan Xu, Qichao Zhang, Songjun Tu, Wei He, Xiangyuan Lan, Xiangyu Tian, Yuqian Fu","submitted_at":"2025-03-17T06:28:25Z","abstract_excerpt":"Recent advancements in post-training methodologies for large language models (LLMs) have highlighted reinforcement learning (RL) as a critical component for enhancing reasoning. However, the substantial computational costs associated with RL-based approaches have led to growing interest in alternative paradigms, such as Direct Preference Optimization (DPO). In this study, we investigate the effectiveness of DPO in facilitating self-improvement for LLMs through iterative preference-based learning. We demonstrate that a single round of DPO with coarse filtering significantly enhances mathematica"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.12854","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/2503.12854/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":"2503.12854","created_at":"2026-07-05T11:44:02.789921+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.12854v3","created_at":"2026-07-05T11:44:02.789921+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.12854","created_at":"2026-07-05T11:44:02.789921+00:00"},{"alias_kind":"pith_short_12","alias_value":"5UTQ3P6RDSIB","created_at":"2026-07-05T11:44:02.789921+00:00"},{"alias_kind":"pith_short_16","alias_value":"5UTQ3P6RDSIBFKAT","created_at":"2026-07-05T11:44:02.789921+00:00"},{"alias_kind":"pith_short_8","alias_value":"5UTQ3P6R","created_at":"2026-07-05T11:44:02.789921+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00869","citing_title":"Enhancing LLM Metacognition via Cognitive Pairwise Training","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23398","citing_title":"TPMM-DPO: Trajectory-aware Preference-guided Model Merging for Iterative Direct Preference Optimization","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21266","citing_title":"How Much Online RL is Enough? Informative Rollouts for Offline Preference Optimization in RLVR","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2508.06412","citing_title":"Sample-efficient LLM Optimization with Reset Replay","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23809","citing_title":"LegalDrill: Diagnosis-Driven Synthesis for Legal Reasoning in Small Language Models","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20933","citing_title":"IRIS: Interpolative R\\'enyi Iterative Self-play for Large Language Model Fine-Tuning","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07353","citing_title":"Confidence-Aware Alignment Makes Reasoning LLMs More Reliable","ref_index":42,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX","json":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX.json","graph_json":"https://pith.science/api/pith-number/5UTQ3P6RDSIBFKATPGBSEW4ZAX/graph.json","events_json":"https://pith.science/api/pith-number/5UTQ3P6RDSIBFKATPGBSEW4ZAX/events.json","paper":"https://pith.science/paper/5UTQ3P6R"},"agent_actions":{"view_html":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX","download_json":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX.json","view_paper":"https://pith.science/paper/5UTQ3P6R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.12854&json=true","fetch_graph":"https://pith.science/api/pith-number/5UTQ3P6RDSIBFKATPGBSEW4ZAX/graph.json","fetch_events":"https://pith.science/api/pith-number/5UTQ3P6RDSIBFKATPGBSEW4ZAX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX/action/storage_attestation","attest_author":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX/action/author_attestation","sign_citation":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX/action/citation_signature","submit_replication":"https://pith.science/pith/5UTQ3P6RDSIBFKATPGBSEW4ZAX/action/replication_record"}},"created_at":"2026-07-05T11:44:02.789921+00:00","updated_at":"2026-07-05T11:44:02.789921+00:00"}