{"id":"https://openalex.org/W4404988001","doi":"https://doi.org/10.48550/arxiv.2411.16646","title":"Self-Generated Critiques Boost Reward Modeling for Language Models","display_name":"Self-Generated Critiques Boost Reward Modeling for Language Models","publication_year":2024,"publication_date":"2024-11-25","ids":{"openalex":"https://openalex.org/W4404988001","doi":"https://doi.org/10.48550/arxiv.2411.16646"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2411.16646","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2411.16646","pdf_url":"https://arxiv.org/pdf/2411.16646","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2411.16646","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100397967","display_name":"Yue Yu","orcid":"https://orcid.org/0000-0002-3683-5208"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Yue","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010176677","display_name":"Zhengxing Chen","orcid":"https://orcid.org/0000-0003-0252-6754"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Zhengxing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079086787","display_name":"Aston Zhang","orcid":"https://orcid.org/0000-0003-0890-0524"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Aston","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103004021","display_name":"Liang Soon Tan","orcid":"https://orcid.org/0000-0002-6057-4378"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tan, Liang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054819831","display_name":"Chenguang Zhu","orcid":"https://orcid.org/0000-0002-7343-8279"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Chenguang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103936332","display_name":"Richard Yuanzhe Pang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pang, Richard Yuanzhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112235930","display_name":"Yundi Qian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qian, Yundi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100684659","display_name":"Xuewei Wang","orcid":"https://orcid.org/0000-0002-2704-3444"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Xuewei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075783850","display_name":"Suchin Gururangan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gururangan, Suchin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115596350","display_name":"Chao Zhang","orcid":"https://orcid.org/0009-0008-2674-344X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Chao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024510591","display_name":"Melanie Kambadur","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kambadur, Melanie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058067505","display_name":"Dhruv Mahajan","orcid":"https://orcid.org/0000-0001-6611-8018"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mahajan, Dhruv","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5114202928","display_name":"Rui Hou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hou, Rui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.982200026512146,"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":0.982200026512146,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9675999879837036,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4753301441669464},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.414465069770813},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.4005941152572632},{"id":"https://openalex.org/keywords/cognitive-science","display_name":"Cognitive science","score":0.3449985980987549},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.2789705991744995}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4753301441669464},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.414465069770813},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.4005941152572632},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.3449985980987549},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2789705991744995}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2411.16646","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2411.16646","pdf_url":"https://arxiv.org/pdf/2411.16646","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2411.16646","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2411.16646","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article-journal"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2411.16646","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2411.16646","pdf_url":"https://arxiv.org/pdf/2411.16646","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4404988001.pdf","grobid_xml":"https://content.openalex.org/works/W4404988001.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Reward":[0],"modeling":[1,52,100],"is":[2],"crucial":[3],"for":[4],"aligning":[5],"large":[6],"language":[7,37],"models":[8,24,65,108],"(LLMs)":[9],"with":[10,132],"human":[11,18],"preferences,":[12],"especially":[13],"in":[14,34,127,135],"reinforcement":[15],"learning":[16],"from":[17],"feedback":[19],"(RLHF).":[20],"However,":[21],"current":[22],"reward":[23,48,51,64,87,99,107],"mainly":[25],"produce":[26],"scalar":[27,47],"scores":[28],"and":[29,45,78,89,109,115],"struggle":[30],"to":[31,105],"incorporate":[32],"critiques":[33,44,68,126],"a":[35,60,74],"natural":[36],"format.":[38],"We":[39],"hypothesize":[40],"that":[41,62,96],"predicting":[42],"both":[43],"the":[46,122],"would":[49],"improve":[50],"ability.":[53],"Motivated":[54],"by":[55,83,102],"this,":[56],"we":[57],"propose":[58],"Critic-RM,":[59],"framework":[61],"improves":[63,98],"using":[66],"self-generated":[67],"without":[69],"extra":[70],"supervision.":[71],"Critic-RM":[72,97],"employs":[73],"two-stage":[75],"process:":[76],"generating":[77],"filtering":[79],"high-quality":[80],"critiques,":[81],"followed":[82],"joint":[84],"fine-tuning":[85],"on":[86],"prediction":[88],"critique":[90],"generation.":[91],"Experiments":[92],"across":[93],"benchmarks":[94],"show":[95],"accuracy":[101],"3.7%-7.3%":[103],"compared":[104],"standard":[106],"LLM":[110],"judges,":[111],"demonstrating":[112],"strong":[113],"performance":[114],"data":[116],"efficiency.":[117],"Additional":[118],"studies":[119],"further":[120],"validate":[121],"effectiveness":[123],"of":[124],"generated":[125],"rectifying":[128],"flawed":[129],"reasoning":[130,137],"steps":[131],"2.5%-3.2%":[133],"gains":[134],"improving":[136],"accuracy.":[138]},"counts_by_year":[],"updated_date":"2026-08-19T15:00:24.278416","created_date":"2025-10-10T00:00:00"}