{"id":"https://openalex.org/W4416035975","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.78","title":"Training Language Models to Critique With Multi-agent Feedback","display_name":"Training Language Models to Critique With Multi-agent Feedback","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416035975","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.78"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.78","pdf_url":"https://aclanthology.org/2025.findings-emnlp.78.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":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-emnlp.78.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5110379714","display_name":"Tian Lan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian Lan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100619686","display_name":"Wenwei Zhang","orcid":"https://orcid.org/0000-0002-2745-9773"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenwei Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067731176","display_name":"Chengqi Lyu","orcid":"https://orcid.org/0000-0002-4438-1314"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chengqi Lyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108313945","display_name":"Shuaibin Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuaibin Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100385737","display_name":"Xu Chen","orcid":"https://orcid.org/0000-0002-8425-036X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087631670","display_name":"Heyan Huang","orcid":"https://orcid.org/0000-0002-0320-7520"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Heyan Huang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010087030","display_name":"Dahua Lin","orcid":"https://orcid.org/0000-0002-8865-7896"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dahua Lin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017626590","display_name":"Xian-Ling Mao","orcid":"https://orcid.org/0000-0001-6795-2311"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xian-Ling Mao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100438001","display_name":"Kai Chen","orcid":"https://orcid.org/0000-0003-2587-6028"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kai Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":12.1274,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.98066823,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1474","last_page":"1501"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12090","display_name":"Language and cultural evolution","score":0.13330000638961792,"subfield":{"id":"https://openalex.org/subfields/3316","display_name":"Cultural Studies"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T12090","display_name":"Language and cultural evolution","score":0.13330000638961792,"subfield":{"id":"https://openalex.org/subfields/3316","display_name":"Cultural Studies"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.07620000094175339,"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/T11883","display_name":"Embodied and Extended Cognition","score":0.05990000069141388,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.5867999792098999},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.25519999861717224},{"id":"https://openalex.org/keywords/language-acquisition","display_name":"Language acquisition","score":0.24629999697208405}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.5867999792098999},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.4657000005245209},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.45080000162124634},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.35409998893737793},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.31769999861717224},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.3059999942779541},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.30300000309944153},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2556999921798706},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.24629999697208405}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.78","pdf_url":"https://aclanthology.org/2025.findings-emnlp.78.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":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.78","pdf_url":"https://aclanthology.org/2025.findings-emnlp.78.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":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2563629645","display_name":null,"funder_award_id":"U21B2009","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G6095425281","display_name":null,"funder_award_id":"62276110","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8878491478","display_name":null,"funder_award_id":"62172039","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321543","display_name":"China Postdoctoral Science Foundation","ror":"https://ror.org/0426zh255"},{"id":"https://openalex.org/F4320323110","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74"},{"id":"https://openalex.org/F4320327514","display_name":"Beijing Institute of Technology Research Fund Program for Young Scholars","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416035975.pdf","grobid_xml":"https://content.openalex.org/works/W4416035975.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Critique":[0],"ability,":[1],"a":[2,35,84,115],"meta-cognitive":[3],"capability":[4,179],"of":[5,58,137,160,180],"humans,":[6],"presents":[7],"significant":[8],"challenges":[9],"for":[10],"LLMs":[11,61,194],"to":[12,55,91,132],"improve.While":[13],"utilizing":[14],"human":[15],"annotation":[16],"can":[17],"enhance":[18],"critique":[19,93,141],"ability":[20,94],"effectively,":[21],"most":[22],"recent":[23],"works":[24],"primarily":[25],"rely":[26],"on":[27,62,143,153,171],"supervised":[28,98],"fine-tuning":[29,60,99],"(SFT)":[30,100],"using":[31],"critiques":[32,48,65,113,123],"generated":[33],"by":[34,128],"single":[36],"LLM":[37],"like":[38],"GPT-4,":[39],"which":[40],"is":[41],"more":[42],"scalable":[43],"and":[44,101,126,148,174,195],"cost-effective.However,":[45],"such":[46],"modelgenerated":[47],"often":[49],"suffer":[50],"from":[51],"inherent":[52],"flaws":[53],"due":[54],"the":[56,75,97,106,119,157,177],"complexity":[57],"critique.Consequently,":[59],"these":[63],"flawed":[64],"not":[66],"only":[67],"limits":[68],"performance":[69],"but":[70],"also":[71],"propagates":[72],"errors":[73],"into":[74],"learned":[76],"model.To":[77],"address":[78],"this":[79],"issue,":[80],"we":[81,145],"propose":[82],"MultiCritique,":[83,144],"unified":[85],"framework":[86],"that":[87],"leverages":[88],"multi-agent":[89],"feedback":[90],"improve":[92],"in":[95,139,176],"both":[96],"reinforcement":[102],"learning":[103],"(RL)":[104],"stages.In":[105],"SFT":[107,147],"stage,":[108,121],"MultiCritique":[109],"aggregates":[110],"high-quality":[111],"multiagent":[112],"through":[114],"fine-grained":[116],"metacritique":[117],"mechanism.In":[118],"RL":[120,138,149],"preference":[122],"are":[124],"constructed":[125],"refined":[127],"validating":[129],"their":[130],"contributions":[131],"revisions,":[133],"thereby":[134],"enhancing":[135],"robustness":[136],"improving":[140],"ability.Based":[142],"construct":[146],"datasets.Extensive":[150],"experimental":[151],"results":[152],"two":[154],"benchmarks":[155],"highlight":[156],"key":[158],"benefits":[159],"our":[161,182],"dataset,":[162],"including":[163],"superior":[164],"quality,":[165],"enhanced":[166],"data":[167],"efficiency,":[168],"strong":[169],"generalization":[170],"unseen":[172],"tasks,":[173],"improvements":[175],"general":[178],"LLMs.Notably,":[181],"fine-tuned":[183],"7B":[184],"model":[185],"significantly":[186],"surpasses":[187],"advanced":[188,192],"7B-13B":[189],"models,":[190],"approaching":[191],"70B":[193],"GPT-4.Resources":[196],"have":[197],"been":[198],"publicly":[199],"available":[200],"1":[201],".":[202]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-11-08T00:00:00"}