{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LWNQYQRP5FO7FWQKIWBZRBTRNK","short_pith_number":"pith:LWNQYQRP","schema_version":"1.0","canonical_sha256":"5d9b0c422fe95df2da0a45839886716aa078564ca3f2a8dfba1eeb8c25066df9","source":{"kind":"arxiv","id":"2409.02392","version":2},"attestation_state":"computed","paper":{"title":"Building Math Agents with Multi-Turn Iterative Preference Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Aviv Rosenberg, Bilal Piot, Chengshuai Shi, Chi Jin, Daniele Calandriello, Jiaming Shen, Misha Khalman, Mohammad Saleh, Rishabh Joshi, Tianqi Liu, Tong Zhang, Wei Xiong, Zhen Qin","submitted_at":"2024-09-04T02:41:04Z","abstract_excerpt":"Recent studies have shown that large language models' (LLMs) mathematical problem-solving capabilities can be enhanced by integrating external tools, such as code interpreters, and employing multi-turn Chain-of-Thought (CoT) reasoning. While current methods focus on synthetic data generation and Supervised Fine-Tuning (SFT), this paper studies the complementary direct preference learning approach to further improve model performance. However, existing direct preference learning algorithms are originally designed for the single-turn chat task, and do not fully address the complexities of multi-"},"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":"2409.02392","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-04T02:41:04Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"db82532ddd50b0ba9d6b121d70296cd5bd38dcba15676993f99cbab9add86189","abstract_canon_sha256":"b4bdca8e408ae9de923a6624d368f022ff58465a5093e87fa53fe942d1dec91b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:21:22.809885Z","signature_b64":"+uuigc9f34znzyaqGAqqKeITCaEjrYC2ybcQnfIfZpdOwMBgWASwJ0uZVJAK96O2LsFpwiMVZKkTtaf2o4xHDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5d9b0c422fe95df2da0a45839886716aa078564ca3f2a8dfba1eeb8c25066df9","last_reissued_at":"2026-07-05T10:21:22.809078Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:21:22.809078Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Building Math Agents with Multi-Turn Iterative Preference Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Aviv Rosenberg, Bilal Piot, Chengshuai Shi, Chi Jin, Daniele Calandriello, Jiaming Shen, Misha Khalman, Mohammad Saleh, Rishabh Joshi, Tianqi Liu, Tong Zhang, Wei Xiong, Zhen Qin","submitted_at":"2024-09-04T02:41:04Z","abstract_excerpt":"Recent studies have shown that large language models' (LLMs) mathematical problem-solving capabilities can be enhanced by integrating external tools, such as code interpreters, and employing multi-turn Chain-of-Thought (CoT) reasoning. While current methods focus on synthetic data generation and Supervised Fine-Tuning (SFT), this paper studies the complementary direct preference learning approach to further improve model performance. However, existing direct preference learning algorithms are originally designed for the single-turn chat task, and do not fully address the complexities of multi-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.02392","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/2409.02392/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":"2409.02392","created_at":"2026-07-05T10:21:22.809192+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.02392v2","created_at":"2026-07-05T10:21:22.809192+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.02392","created_at":"2026-07-05T10:21:22.809192+00:00"},{"alias_kind":"pith_short_12","alias_value":"LWNQYQRP5FO7","created_at":"2026-07-05T10:21:22.809192+00:00"},{"alias_kind":"pith_short_16","alias_value":"LWNQYQRP5FO7FWQK","created_at":"2026-07-05T10:21:22.809192+00:00"},{"alias_kind":"pith_short_8","alias_value":"LWNQYQRP","created_at":"2026-07-05T10:21:22.809192+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18910","citing_title":"REVES: REvision and VErification--Augmented Training for Test-Time Scaling","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2509.03403","citing_title":"Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2409.12917","citing_title":"Training Language Models to Self-Correct via Reinforcement Learning","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK","json":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK.json","graph_json":"https://pith.science/api/pith-number/LWNQYQRP5FO7FWQKIWBZRBTRNK/graph.json","events_json":"https://pith.science/api/pith-number/LWNQYQRP5FO7FWQKIWBZRBTRNK/events.json","paper":"https://pith.science/paper/LWNQYQRP"},"agent_actions":{"view_html":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK","download_json":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK.json","view_paper":"https://pith.science/paper/LWNQYQRP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.02392&json=true","fetch_graph":"https://pith.science/api/pith-number/LWNQYQRP5FO7FWQKIWBZRBTRNK/graph.json","fetch_events":"https://pith.science/api/pith-number/LWNQYQRP5FO7FWQKIWBZRBTRNK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK/action/storage_attestation","attest_author":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK/action/author_attestation","sign_citation":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK/action/citation_signature","submit_replication":"https://pith.science/pith/LWNQYQRP5FO7FWQKIWBZRBTRNK/action/replication_record"}},"created_at":"2026-07-05T10:21:22.809192+00:00","updated_at":"2026-07-05T10:21:22.809192+00:00"}