{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3ISFIZ7QIOMPS3WQU3N7TMG5KC","short_pith_number":"pith:3ISFIZ7Q","schema_version":"1.0","canonical_sha256":"da245467f04398f96ed0a6dbf9b0dd50bad7df6e43cb9cfb9ea529d6ca54e2d2","source":{"kind":"arxiv","id":"2406.10858","version":2},"attestation_state":"computed","paper":{"title":"Step-level Value Preference Optimization for Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chengxi Li, Guoxin Chen, Kai Fan, Minpeng Liao","submitted_at":"2024-06-16T09:06:17Z","abstract_excerpt":"Direct Preference Optimization (DPO) using an implicit reward model has proven to be an effective alternative to reinforcement learning from human feedback (RLHF) for fine-tuning preference aligned large language models (LLMs). However, the overall preference annotations of responses do not fully capture the fine-grained quality of model outputs in complex multi-step reasoning tasks, such as mathematical reasoning. To address this limitation, we introduce a novel algorithm called Step-level Value Preference Optimization (SVPO). Our approach employs Monte Carlo Tree Search (MCTS) to automatical"},"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":"2406.10858","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-16T09:06:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"40bc6e6e4306b05eeefed9528720c4dcc8242358ae99308a7be80fabc9a0deb4","abstract_canon_sha256":"fdf67f015bcae8d98410a63583a3933e61b5111c58b209871caca7608ff74b67"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:12:29.015482Z","signature_b64":"qukGFD8763m/qVbAhEjroYoo1liq3c/OcOyhXZUyb+0GJ7U+xbk6Gb9cMD00FnAFWxrFjHsbZGwBppaqnb/6AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"da245467f04398f96ed0a6dbf9b0dd50bad7df6e43cb9cfb9ea529d6ca54e2d2","last_reissued_at":"2026-07-05T09:12:29.014980Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:12:29.014980Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Step-level Value Preference Optimization for Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chengxi Li, Guoxin Chen, Kai Fan, Minpeng Liao","submitted_at":"2024-06-16T09:06:17Z","abstract_excerpt":"Direct Preference Optimization (DPO) using an implicit reward model has proven to be an effective alternative to reinforcement learning from human feedback (RLHF) for fine-tuning preference aligned large language models (LLMs). However, the overall preference annotations of responses do not fully capture the fine-grained quality of model outputs in complex multi-step reasoning tasks, such as mathematical reasoning. To address this limitation, we introduce a novel algorithm called Step-level Value Preference Optimization (SVPO). Our approach employs Monte Carlo Tree Search (MCTS) to automatical"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.10858","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/2406.10858/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":"2406.10858","created_at":"2026-07-05T09:12:29.015038+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.10858v2","created_at":"2026-07-05T09:12:29.015038+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.10858","created_at":"2026-07-05T09:12:29.015038+00:00"},{"alias_kind":"pith_short_12","alias_value":"3ISFIZ7QIOMP","created_at":"2026-07-05T09:12:29.015038+00:00"},{"alias_kind":"pith_short_16","alias_value":"3ISFIZ7QIOMPS3WQ","created_at":"2026-07-05T09:12:29.015038+00:00"},{"alias_kind":"pith_short_8","alias_value":"3ISFIZ7Q","created_at":"2026-07-05T09:12:29.015038+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2408.07199","citing_title":"Agent Q: Advanced Reasoning and Learning for Autonomous AI Agents","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04595","citing_title":"SpikingMamba: Towards Energy-Efficient Large Language Models via Knowledge Distillation from Mamba","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2510.10649","citing_title":"Unlocking Exploration in RLVR: Uncertainty-aware Advantage Shaping for Deeper Reasoning","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2503.12937","citing_title":"R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimization","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2501.09686","citing_title":"Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2412.21187","citing_title":"Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11974","citing_title":"Towards Order Fairness: Mitigating LLMs Order Sensitivity through Dual Group Advantage Optimization","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09492","citing_title":"APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC","json":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC.json","graph_json":"https://pith.science/api/pith-number/3ISFIZ7QIOMPS3WQU3N7TMG5KC/graph.json","events_json":"https://pith.science/api/pith-number/3ISFIZ7QIOMPS3WQU3N7TMG5KC/events.json","paper":"https://pith.science/paper/3ISFIZ7Q"},"agent_actions":{"view_html":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC","download_json":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC.json","view_paper":"https://pith.science/paper/3ISFIZ7Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.10858&json=true","fetch_graph":"https://pith.science/api/pith-number/3ISFIZ7QIOMPS3WQU3N7TMG5KC/graph.json","fetch_events":"https://pith.science/api/pith-number/3ISFIZ7QIOMPS3WQU3N7TMG5KC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC/action/storage_attestation","attest_author":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC/action/author_attestation","sign_citation":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC/action/citation_signature","submit_replication":"https://pith.science/pith/3ISFIZ7QIOMPS3WQU3N7TMG5KC/action/replication_record"}},"created_at":"2026-07-05T09:12:29.015038+00:00","updated_at":"2026-07-05T09:12:29.015038+00:00"}