{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:G4F5OO4YXFUCNXVAQRTIGUDQAC","short_pith_number":"pith:G4F5OO4Y","schema_version":"1.0","canonical_sha256":"370bd73b98b96826dea0846683507000bc7f2b6f936ecb3b0903cddde45723cf","source":{"kind":"arxiv","id":"2606.21943","version":1},"attestation_state":"computed","paper":{"title":"Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Annie Wong, Aske Plaat, Chao Gao, Filip Ilievski, Hengyuan Zhang, Jacob E. Kooi, Jiayang Shi, Kevin Qiu, Lincen Yang, Mark Hoogendoorn, Ngai Wong, Qi Huang, Shiping Yang, Shujian Yu, Ting-Chih Chen, Vincent Fran\\c{c}ois-Lavet, Xinrui Zu, Yuxuan Jiang, Zhaochun Ren, Zhao Yang, Zhong Li","submitted_at":"2026-06-20T08:20:41Z","abstract_excerpt":"Reinforcement learning (RL) has become central to LLM post-training, yet the methods that dominate current pipelines, PPO and GRPO, represent only a narrow slice of what RL offers. Understanding why these methods prevail, and what alternatives exist, requires a principled examination of the design decisions that underlie any RL algorithm.\n  This survey organizes that examination around three stages of algorithm construction. We begin with MDP creation: how the reward function, state space, action space, termination condition, and discount factor are, or could be, defined for LLM training. We t"},"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":"2606.21943","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-20T08:20:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6c73c6288d7418e1629c69ea6e0ab32845916f08a5dc94c6bfbc9d1ffe94bad6","abstract_canon_sha256":"d94d6465d99fa3adb141f952e9a72fcf97f44cee182aa975873790cc219214f2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T02:13:03.789659Z","signature_b64":"F2Gv9TOk4J7XDTMrn4pC+8NrjYV4S9kC0Rh4r3lZD8yOj3jbXkmfHTVIb1YxwnK+Lx2U4YlN58G7uGEF6mIvDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"370bd73b98b96826dea0846683507000bc7f2b6f936ecb3b0903cddde45723cf","last_reissued_at":"2026-06-23T02:13:03.789242Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T02:13:03.789242Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Annie Wong, Aske Plaat, Chao Gao, Filip Ilievski, Hengyuan Zhang, Jacob E. Kooi, Jiayang Shi, Kevin Qiu, Lincen Yang, Mark Hoogendoorn, Ngai Wong, Qi Huang, Shiping Yang, Shujian Yu, Ting-Chih Chen, Vincent Fran\\c{c}ois-Lavet, Xinrui Zu, Yuxuan Jiang, Zhaochun Ren, Zhao Yang, Zhong Li","submitted_at":"2026-06-20T08:20:41Z","abstract_excerpt":"Reinforcement learning (RL) has become central to LLM post-training, yet the methods that dominate current pipelines, PPO and GRPO, represent only a narrow slice of what RL offers. Understanding why these methods prevail, and what alternatives exist, requires a principled examination of the design decisions that underlie any RL algorithm.\n  This survey organizes that examination around three stages of algorithm construction. We begin with MDP creation: how the reward function, state space, action space, termination condition, and discount factor are, or could be, defined for LLM training. We t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.21943","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/2606.21943/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":"2606.21943","created_at":"2026-06-23T02:13:03.789304+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.21943v1","created_at":"2026-06-23T02:13:03.789304+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.21943","created_at":"2026-06-23T02:13:03.789304+00:00"},{"alias_kind":"pith_short_12","alias_value":"G4F5OO4YXFUC","created_at":"2026-06-23T02:13:03.789304+00:00"},{"alias_kind":"pith_short_16","alias_value":"G4F5OO4YXFUCNXVA","created_at":"2026-06-23T02:13:03.789304+00:00"},{"alias_kind":"pith_short_8","alias_value":"G4F5OO4Y","created_at":"2026-06-23T02:13:03.789304+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC","json":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC.json","graph_json":"https://pith.science/api/pith-number/G4F5OO4YXFUCNXVAQRTIGUDQAC/graph.json","events_json":"https://pith.science/api/pith-number/G4F5OO4YXFUCNXVAQRTIGUDQAC/events.json","paper":"https://pith.science/paper/G4F5OO4Y"},"agent_actions":{"view_html":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC","download_json":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC.json","view_paper":"https://pith.science/paper/G4F5OO4Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.21943&json=true","fetch_graph":"https://pith.science/api/pith-number/G4F5OO4YXFUCNXVAQRTIGUDQAC/graph.json","fetch_events":"https://pith.science/api/pith-number/G4F5OO4YXFUCNXVAQRTIGUDQAC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC/action/storage_attestation","attest_author":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC/action/author_attestation","sign_citation":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC/action/citation_signature","submit_replication":"https://pith.science/pith/G4F5OO4YXFUCNXVAQRTIGUDQAC/action/replication_record"}},"created_at":"2026-06-23T02:13:03.789304+00:00","updated_at":"2026-06-23T02:13:03.789304+00:00"}