{"id":"https://openalex.org/W7138054462","doi":"https://doi.org/10.1609/aaai.v40i36.40341","title":"Group Causal Policy Optimization for Post-Training Large Language Models","display_name":"Group Causal Policy Optimization for Post-Training Large Language Models","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138054462","doi":"https://doi.org/10.1609/aaai.v40i36.40341"},"language":"en","primary_location":{"id":"doi:10.1609/aaai.v40i36.40341","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40341","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40341/44302","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40341/44302","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5011058446","display_name":"Ziyin Gu","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ziyin Gu","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129727961","display_name":"Jingyao Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jingyao Wang","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068072183","display_name":"Ran Zuo","orcid":"https://orcid.org/0000-0002-9746-3487"},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ran Zuo","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences,\nCommunication University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences,\nCommunication University of China","institution_ids":["https://openalex.org/I4210128818","https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129644910","display_name":"Chuxiong Sun","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chuxiong Sun","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129677282","display_name":"Zeen Song","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeen Song","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129727938","display_name":"Changwen Zheng","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Changwen Zheng","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5129739126","display_name":"Wenwen Qiang","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128818","display_name":"Institute of Software","ror":"https://ror.org/033dfsn42","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210128818"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wenwen Qiang","raw_affiliation_strings":["Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Software Chinese Academy of Sciences,\nUniversity of the Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210128818"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":20.9207,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.97705352,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"40","issue":"36","first_page":"30834","last_page":"30842"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2669999897480011,"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/T10028","display_name":"Topic Modeling","score":0.2669999897480011,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.07339999824762344,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.05950000137090683,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/complementarity","display_name":"Complementarity (molecular biology)","score":0.7240999937057495},{"id":"https://openalex.org/keywords/subspace-topology","display_name":"Subspace topology","score":0.5095999836921692},{"id":"https://openalex.org/keywords/causal-model","display_name":"Causal model","score":0.4700999855995178},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4611999988555908},{"id":"https://openalex.org/keywords/causal-structure","display_name":"Causal structure","score":0.4496000111103058},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.44699999690055847},{"id":"https://openalex.org/keywords/function","display_name":"Function (biology)","score":0.4339999854564667},{"id":"https://openalex.org/keywords/group","display_name":"Group (periodic table)","score":0.42419999837875366}],"concepts":[{"id":"https://openalex.org/C202269582","wikidata":"https://www.wikidata.org/wiki/Q2644277","display_name":"Complementarity (molecular biology)","level":2,"score":0.7240999937057495},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6298999786376953},{"id":"https://openalex.org/C32834561","wikidata":"https://www.wikidata.org/wiki/Q660730","display_name":"Subspace topology","level":2,"score":0.5095999836921692},{"id":"https://openalex.org/C11671645","wikidata":"https://www.wikidata.org/wiki/Q5054567","display_name":"Causal model","level":2,"score":0.4700999855995178},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4611999988555908},{"id":"https://openalex.org/C163504300","wikidata":"https://www.wikidata.org/wiki/Q2364925","display_name":"Causal structure","level":2,"score":0.4496000111103058},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.44699999690055847},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.4339999854564667},{"id":"https://openalex.org/C2781311116","wikidata":"https://www.wikidata.org/wiki/Q83306","display_name":"Group (periodic table)","level":2,"score":0.42419999837875366},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41850000619888306},{"id":"https://openalex.org/C57493831","wikidata":"https://www.wikidata.org/wiki/Q3134666","display_name":"Projection (relational algebra)","level":2,"score":0.38609999418258667},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.3635999858379364},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.34700000286102295},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3458999991416931},{"id":"https://openalex.org/C65660741","wikidata":"https://www.wikidata.org/wiki/Q3952743","display_name":"Score","level":2,"score":0.32850000262260437},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3208000063896179},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.27320000529289246},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.25920000672340393}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1609/aaai.v40i36.40341","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40341","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40341/44302","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},{"id":"pmh:oai:ojs.aaai.org:article/40341","is_oa":false,"landing_page_url":"https://ojs.aaai.org/index.php/AAAI/article/view/40341","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2159-5399","raw_type":"info:eu-repo/semantics/article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i36.40341","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40341","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40341/44302","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-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/F4320332382","display_name":"Oak Ridge Institute for Science and Education","ror":"https://ror.org/0526p1y61"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7138054462.pdf","grobid_xml":"https://content.openalex.org/works/W7138054462.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"advances":[1],"in":[2],"large":[3],"language":[4],"models":[5],"(LLMs)":[6],"have":[7],"broadened":[8],"their":[9],"applicability":[10],"across":[11],"diverse":[12],"tasks,":[13],"yet":[14],"specialized":[15],"domains":[16],"still":[17],"require":[18],"targeted":[19],"post-training.":[20],"Among":[21],"existing":[22,169],"methods,":[23],"Group":[24,125],"Relative":[25],"Policy":[26,127],"Optimization":[27,128],"(GRPO)":[28],"stands":[29],"out":[30],"for":[31],"its":[32],"efficiency,":[33],"leveraging":[34],"groupwise":[35],"relative":[36],"rewards":[37],"while":[38],"avoiding":[39],"costly":[40],"value":[41],"function":[42],"learning.":[43],"However,":[44],"GRPO":[45],"treats":[46],"candidate":[47,76],"responses":[48,77,100],"as":[49,55],"independent,":[50],"overlooking":[51],"semantic":[52],"interactions":[53],"such":[54],"complementarity":[56],"and":[57,108,144],"contradiction.":[58],"To":[59],"address":[60],"this":[61,110],"challenge,":[62],"we":[63,123],"first":[64],"introduce":[65],"a":[66,82,87,102,113,140,145,154],"Structural":[67],"Causal":[68,126],"Model":[69],"(SCM)":[70],"that":[71,149,165],"reveals":[72],"hidden":[73],"dependencies":[74],"among":[75],"induced":[78],"by":[79],"conditioning":[80],"on":[81,120,161],"final":[83],"integrated":[84],"output,":[85],"forming":[86],"collider":[88],"structure.":[89],"Then,":[90],"our":[91],"causal":[92,132],"analysis":[93],"leads":[94],"to":[95],"two":[96,137],"insights:":[97],"(1)":[98],"projecting":[99],"onto":[101],"causally-informed":[103,141],"subspace":[104],"improves":[105],"prediction":[106],"quality,":[107],"(2)":[109],"projection":[111],"yields":[112],"better":[114],"baseline":[115],"than":[116],"query-only":[117],"conditioning.":[118],"Building":[119],"these":[121],"insights,":[122],"propose":[124],"(GCPO),":[129],"which":[130],"integrates":[131],"structure":[133],"into":[134],"optimization":[135],"through":[136],"key":[138],"components:":[139],"reward":[142],"adjustment":[143],"novel":[146],"KL-regularization":[147],"term":[148],"aligns":[150],"the":[151],"policy":[152],"with":[153],"causally-projected":[155],"reference":[156],"distribution.":[157],"Comprehensive":[158],"experimental":[159],"evaluations":[160],"various":[162],"benchmarks":[163],"demonstrate":[164],"GCPO":[166],"consistently":[167],"surpasses":[168],"methods.":[170]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2026-03-18T00:00:00"}