{"id":"https://openalex.org/W4412945254","doi":"https://doi.org/10.18653/v1/2025.acl-long.604","title":"TreeRL: LLM Reinforcement Learning with On-Policy Tree Search","display_name":"TreeRL: LLM Reinforcement Learning with On-Policy Tree Search","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945254","doi":"https://doi.org/10.18653/v1/2025.acl-long.604"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.604","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.604","pdf_url":"https://aclanthology.org/2025.acl-long.604.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.604.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5035898033","display_name":"Zhenyu Hou","orcid":"https://orcid.org/0000-0003-2036-0778"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhenyu Hou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060639952","display_name":"Ziniu Hu","orcid":"https://orcid.org/0009-0007-8818-739X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ziniu Hu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034694224","display_name":"Yujiang Li","orcid":"https://orcid.org/0000-0002-2008-223X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yujiang Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112111778","display_name":"Rui Lu","orcid":"https://orcid.org/0009-0003-4850-8401"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rui Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100670350","display_name":"Jie Tang","orcid":"https://orcid.org/0000-0001-8602-7754"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jie Tang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5052284218","display_name":"Yuxiao Dong","orcid":"https://orcid.org/0000-0002-6092-2002"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuxiao Dong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.7588,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.87860697,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"12355","last_page":"12369"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14351","display_name":"Statistical and Computational Modeling","score":0.7240999937057495,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T14351","display_name":"Statistical and Computational Modeling","score":0.7240999937057495,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.7067999839782715,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13999","display_name":"Digital Rights Management and Security","score":0.6920999884605408,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.8188968300819397},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6740570068359375},{"id":"https://openalex.org/keywords/tree","display_name":"Tree (set theory)","score":0.5712238550186157},{"id":"https://openalex.org/keywords/policy-learning","display_name":"Policy learning","score":0.5048940777778625},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.44226527214050293},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.40035367012023926},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.09770852327346802}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.8188968300819397},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6740570068359375},{"id":"https://openalex.org/C113174947","wikidata":"https://www.wikidata.org/wiki/Q2859736","display_name":"Tree (set theory)","level":2,"score":0.5712238550186157},{"id":"https://openalex.org/C2779436431","wikidata":"https://www.wikidata.org/wiki/Q30672407","display_name":"Policy learning","level":2,"score":0.5048940777778625},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44226527214050293},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.40035367012023926},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.09770852327346802},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.604","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.604","pdf_url":"https://aclanthology.org/2025.acl-long.604.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.604","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.604","pdf_url":"https://aclanthology.org/2025.acl-long.604.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320322392","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945254.pdf","grobid_xml":"https://content.openalex.org/works/W4412945254.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2961085424","https://openalex.org/W4306674287","https://openalex.org/W4387369504","https://openalex.org/W4394896187","https://openalex.org/W3170094116","https://openalex.org/W4386462264","https://openalex.org/W3107602296","https://openalex.org/W4364306694","https://openalex.org/W4312192474","https://openalex.org/W4283697347"],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1,52],"(RL)":[2],"with":[3,20],"tree":[4,23,58,97,144],"search":[5,24,59,98,103,145],"has":[6],"demonstrated":[7],"superior":[8,134],"performance":[9,135],"in":[10,44],"traditional":[11,138],"reasoning":[12,30,128],"tasks.Compared":[13],"to":[14,137],"conventional":[15],"independent":[16],"chain":[17],"sampling":[18],"strategies":[19],"outcome":[21],"supervision,":[22],"enables":[25],"better":[26],"exploration":[27],"of":[28,143],"the":[29,69,106,141],"space":[31],"and":[32,67,90,126],"provides":[33],"dense,":[34],"on-policy":[35,57],"process":[36,81],"rewards":[37],"during":[38],"RL":[39,61],"training":[40],"but":[41],"remains":[42],"under-explored":[43],"On-Policy":[45],"LLM":[46],"RL.We":[47],"propose":[48],"TreeRL,":[49],"a":[50,79,95],"reinforcement":[51],"framework":[53],"that":[54,100,131],"directly":[55],"incorporates":[56],"for":[60,71,146],"training.Our":[62],"approach":[63,99],"includes":[64],"intermediate":[65,116],"supervision":[66],"eliminates":[68],"need":[70],"separate":[72,80],"reward":[73,82,91],"model":[74],"training.Existing":[75],"approaches":[76],"typically":[77],"train":[78],"model,":[83],"which":[84],"can":[85],"suffer":[86],"from":[87,114],"distribution":[88],"mismatch":[89],"hacking.We":[92],"also":[93],"introduce":[94],"cost-effective":[96],"achieves":[101,133],"higher":[102],"efficiency":[104],"under":[105],"same":[107],"generation":[108],"token":[109],"budget":[110],"by":[111],"strategically":[112],"branching":[113],"high-uncertainty":[115],"steps":[117],"rather":[118],"than":[119],"using":[120],"random":[121],"branching.Experiments":[122],"on":[123],"challenging":[124],"math":[125],"code":[127],"benchmarks":[129],"demonstrate":[130],"TreeRL":[132],"compared":[136],"ChainRL,":[139],"highlighting":[140],"potential":[142],"LLM.TreeRL":[147],"is":[148],"open-sourced":[149],"at":[150],"https://github.com/THUDM/TreeRL.":[151]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}