{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "yu-et-al-dapo-policy-training-2025", "title": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system", "summary": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline.", "boundary": "PPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.", "area": "Reasoning-model reinforcement-learning algorithms", "topic": "Artificial intelligence", "year": 2025, "lane": "HUMAN", "result": "ALGORITHM", "package_id": "yu-et-al-dapo-policy-training-2025", "target": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "humans": [ "Qiying Yu", "Zheng Zhang", "Ruofei Zhu", "Yufeng Yuan", "Xiaochen Zuo", "Yu Yue", "Tiantian Fan", "Gaohong Liu", "Lingjun Liu", "Xin Liu", "Haibin Lin", "Zhiqi Lin", "Bole Ma", "Guangming Sheng", "Yuxuan Tong", "Chi Zhang", "Mofan Zhang", "Wang Zhang", "Hang Zhu", "Jinhua Zhu", "Jiaze Chen", "Jiangjie Chen", "Chengyi Wang", "Hongli Yu", "Weinan Dai", "Yuxuan Song", "Xiangpeng Wei", "Hao Zhou", "Jingjing Liu", "Wei-Ying Ma", "Ya-Qin Zhang", "Lin Yan", "Mu Qiao", "Yonghui Wu", "Mingxuan Wang" ], "authors_complete": true, "human_role_note": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.", "agents": [], "related": [], "subresults": [], "scientific_assessment": { "source_review": "SOURCE_CURATED", "scientific_review": "NOT_REVIEWED_BY_WRS", "formal_replay": "NOT_RUN", "provider_attestation": "NONE" }, "sources": [ { "id": "cur20261015-arxiv-2503-14476v1", "title": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" }, { "id": "cur20261015-support-yu-et-al-dapo-policy-training-2025", "title": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system — author/institution implementation", "url": "https://dapo-sia.github.io/" } ], "points": "130000", "published_grade": { "id": "yu-et-al-dapo-policy-training-2025", "entry_id": "yu-et-al-dapo-policy-training-2025", "band": "A", "modifier": null, "target": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "what_changed": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline.", "scope": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline. PPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.", "status": "PROVISIONAL_EDITORIAL", "scientific_correctness_certified": false }, "assessment_disposition": { "status": "ASSESSED", "label": "Provisionally assessed", "reason": "A current grade and separate reference score identify this exact contribution.", "target": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "edition": "ratings-e039106f30dec177" } }, "version": { "kind": "PUBLIC_RECORD_VIEW", "previous": null, "published_at": "2026-10-16T00:00:00Z", "publisher": { "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33", "role": "WRS_EDITORIAL_PUBLICATION" }, "record": { "contributors": [ { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Qiying Yu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zheng Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ruofei Zhu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yufeng Yuan", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xiaochen Zuo", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yu Yue", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Tiantian Fan", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Gaohong Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Lingjun Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xin Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Haibin Lin", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zhiqi Lin", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Bole Ma", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Guangming Sheng", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yuxuan Tong", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Chi Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Mofan Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Wang Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Hang Zhu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jinhua Zhu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jiaze Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jiangjie Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Chengyi Wang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Hongli Yu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Weinan Dai", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yuxuan Song", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xiangpeng Wei", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Hao Zhou", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jingjing Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Wei-Ying Ma", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ya-Qin Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" } }, { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-1", "kind": "source_reference", "source": { "citation": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system — author/institution implementation", "url": "https://dapo-sia.github.io/" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "PPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.", "domain": "Reasoning-model reinforcement-learning algorithms" }, "sources": [ { "citation": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1" }, { "citation": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system — author/institution implementation", "url": "https://dapo-sia.github.io/" } ], "statement": "# Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system\n\nCombined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline.\n\n## Scientific scope\n\nPPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nQiying Yu; Zheng Zhang; Ruofei Zhu; Yufeng Yuan; Xiaochen Zuo; Yu Yue; Tiantian Fan; Gaohong Liu; Lingjun Liu; Xin Liu; Haibin Lin; Zhiqi Lin; Bole Ma; Guangming Sheng; Yuxuan Tong; Chi Zhang; Mofan Zhang; Wang Zhang; Hang Zhu; Jinhua Zhu; Jiaze Chen; Jiangjie Chen; Chengyi Wang; Hongli Yu; Weinan Dai; Yuxuan Song; Xiangpeng Wei; Hao Zhou; Jingjing Liu; Wei-Ying Ma; Ya-Qin Zhang; Lin Yan; Mu Qiao; Yonghui Wu; Mingxuan Wang\n\nSource-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.\n\n\n\n## Original sources\n\n- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://arxiv.org/abs/2503.14476v1)\n- [Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system — author/institution implementation](https://dapo-sia.github.io/)\n", "summary": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline.", "title": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system" }, "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "source_ref": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:e101eef16be75ff6fa5be83873f72cb0525b917463a5a143a9722ef6f6e9d836" }, "target": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "yu-et-al-dapo-policy-training-2025", "entry_id": "yu-et-al-dapo-policy-training-2025", "revision": 1, "title": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system", "context": "Reasoning-model reinforcement-learning algorithms", "track": "Human-designed research method or resource", "band": "A", "modifier": null, "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-16", "evidence_cutoff": "2026-10-16", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline.", "scope": "Combined asymmetric probability-ratio clipping, filtering of zero-advantage prompt groups and token-weighted policy optimization in an openly documented large-scale reasoning-model RL pipeline. PPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.", "evidence_summary": "The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. PPO, GRPO and verl are inherited. The filtering changes the sampled prompt distribution; source-reported benchmark gains do not prove universal superiority or general safety. Thirty-five unique authors are taken from the explicitly designated Contributions author list, excluding acknowledgments.", "references": [ { "title": "DAPO: An Open-Source LLM Reinforcement Learning System at Scale", "url": "https://arxiv.org/abs/2503.14476v1", "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification." }, { "title": "Yu et al.: DAPO decoupled clipping and dynamic-sampling RL system — author/institution implementation", "url": "https://dapo-sia.github.io/", "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification." } ], "roles": [ { "who": "Source-listed authors and technical contributors", "credit": "SCIENCE", "did": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities." } ], "target": { "record_id": "urn:uuid:d49a9d1e-6e68-4272-ab81-cb6783ddec6b", "version": 1, "content_hash": "sha256:c4c470c02cf0a7a3d012917e73269846fd4f3f49890d91768e2e557bff54dc2a" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "130000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": null, "reading_status": "SOURCE_LINKED_NOTE" }