{"id":"https://openalex.org/W4403579656","doi":"https://doi.org/10.48550/arxiv.2410.13413","title":"Think Thrice Before You Act: Progressive Thought Refinement in Large Language Models","display_name":"Think Thrice Before You Act: Progressive Thought Refinement in Large Language Models","publication_year":2024,"publication_date":"2024-10-17","ids":{"openalex":"https://openalex.org/W4403579656","doi":"https://doi.org/10.48550/arxiv.2410.13413"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2410.13413","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.13413","pdf_url":"https://arxiv.org/pdf/2410.13413","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2410.13413","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5035585332","display_name":"Chengyu Du","orcid":"https://orcid.org/0009-0005-0333-6073","observed_orcids":["https://orcid.org/0009-0005-0333-6073"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Chengyu","raw_affiliation_strings":["Normal University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Normal University","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Han, Jinyi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Jinyi","raw_affiliation_strings":["Shanghai Institute of Artificial Intelligence for Education , East China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Institute of Artificial Intelligence for Education , East China","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ying, Yizhou","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ying, Yizhou","raw_affiliation_strings":["Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063747420","display_name":"Aili Chen","orcid":"https://orcid.org/0000-0002-1751-1208","observed_orcids":["https://orcid.org/0000-0002-1751-1208"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Aili","raw_affiliation_strings":["Normal University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Normal University","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040655519","display_name":"Qianyu He","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"He, Qianyu","raw_affiliation_strings":["Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111174926","display_name":"Haokun Zhao","orcid":"https://orcid.org/0009-0006-5069-8840","observed_orcids":["https://orcid.org/0009-0006-5069-8840"]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhao, Haokun","raw_affiliation_strings":["Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113280402","display_name":"Sirui Xia","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xia, Sirui","raw_affiliation_strings":["Normal University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Normal University","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006747820","display_name":"Haoran Guo","orcid":"https://orcid.org/0009-0001-2205-7601","observed_orcids":["https://orcid.org/0009-0001-2205-7601","https://orcid.org/0009-0008-7407-2740"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Haoran","raw_affiliation_strings":["RhineAI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"RhineAI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075507821","display_name":"Jiaqing Liang","orcid":"https://orcid.org/0000-0003-0670-5602","observed_orcids":["https://orcid.org/0000-0003-0670-5602"]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liang, Jiaqing","raw_affiliation_strings":["School of Data Science , Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Data Science , Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037457961","display_name":"Zulong Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Zulong","raw_affiliation_strings":["Alibaba Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alibaba Group","institution_ids":["https://openalex.org/I45928872"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014174145","display_name":"Liangyue Li","orcid":"https://orcid.org/0000-0001-7630-8851","observed_orcids":["https://orcid.org/0000-0001-7630-8851"]},"institutions":[{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li, Liangyue","raw_affiliation_strings":["Alibaba Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alibaba Group","institution_ids":["https://openalex.org/I45928872"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5115595421","display_name":"Yanghua Xiao","orcid":"https://orcid.org/0000-0001-7417-0799","observed_orcids":["https://orcid.org/0000-0001-7417-0799"]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiao, Yanghua","raw_affiliation_strings":["Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Key Laboratory of Data Science , School of Computer Science , Fudan University","institution_ids":["https://openalex.org/I24943067"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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":0.9394999742507935,"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":0.9394999742507935,"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":0.03350000083446503,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.011900000274181366,"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/large-language-models","display_name":"large language models","score":0.9829999804496765},{"id":"https://openalex.org/keywords/loss-weighting","display_name":"loss weighting","score":0.4390000104904175}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.43965667486190796},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.41841864585876465},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.3401023745536804},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.3257647752761841},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.32285094261169434},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.22725340723991394}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2410.13413","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.13413","pdf_url":"https://arxiv.org/pdf/2410.13413","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2410.13413","doi":"https://doi.org/10.48550/arxiv.2410.13413","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.13413","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,"listed_in":[],"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":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2410.13413","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.13413","pdf_url":"https://arxiv.org/pdf/2410.13413","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4403579656.pdf","grobid_xml":"https://content.openalex.org/works/W4403579656.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4388718076","https://openalex.org/W4414491952","https://openalex.org/W4405766637","https://openalex.org/W4403851177","https://openalex.org/W4407385116","https://openalex.org/W7127364041","https://openalex.org/W7133236450","https://openalex.org/W4387323597","https://openalex.org/W7108211030","https://openalex.org/W4366565380"],"abstract_inverted_index":{"Recent":[0],"advancements":[1],"in":[2,19,45,88,127,189,198],"large":[3],"language":[4],"models":[5],"(LLMs)":[6],"have":[7],"demonstrated":[8],"that":[9,78,170,207],"progressive":[10,110],"refinement,":[11],"rather":[12,162],"than":[13,163],"providing":[14],"a":[15,76,98,108,136],"single":[16],"answer,":[17],"results":[18,168],"more":[20,46,190],"accurate":[21],"and":[22,100,121,143],"thoughtful":[23],"outputs.":[24],"However,":[25],"existing":[26],"methods":[27,52],"often":[28],"rely":[29],"heavily":[30],"on":[31],"supervision":[32],"signals":[33],"to":[34,41,63,81,106,113,119,139,147,150,156,160,183,212],"evaluate":[35],"previous":[36],"responses,":[37],"making":[38],"it":[39],"difficult":[40],"assess":[42],"output":[43],"quality":[44,200],"open-ended":[47,191],"scenarios":[48],"effectively.":[49],"Additionally,":[50],"these":[51,68],"are":[53,124],"typically":[54],"designed":[55],"for":[56],"specific":[57],"tasks,":[58,192],"which":[59],"limits":[60],"their":[61,83],"generalization":[62],"new":[64],"domains.":[65],"To":[66],"address":[67],"limitations,":[69],"we":[70],"propose":[71,97],"Progressive":[72],"Thought":[73,92],"Refinement":[74],"(PTR),":[75],"framework":[77],"enables":[79],"LLMs":[80,149,193,211],"refine":[82,151],"responses":[84,202],"progressively.":[85],"PTR":[86,171,208],"operates":[87],"two":[89],"phases:":[90],"(1)":[91],"data":[93],"construction":[94],"stage:":[95],"We":[96,134],"weak":[99],"strong":[101],"model":[102],"collaborative":[103],"selection":[104],"strategy":[105],"build":[107],"high-quality":[109],"refinement":[111],"dataset":[112],"ensure":[114],"logical":[115],"consistency":[116],"from":[117,181],"thought":[118],"answers,":[120],"the":[122,141,199],"answers":[123],"gradually":[125],"refined":[126],"each":[128],"round.":[129],"(2)":[130],"Thought-Mask":[131],"Fine-Tuning":[132],"Phase:":[133],"design":[135],"training":[137],"structure":[138],"mask":[140],"\"thought\"":[142],"adjust":[144],"loss":[145],"weights":[146],"encourage":[148],"prior":[152],"thought,":[153],"teaching":[154],"them":[155],"implicitly":[157],"understand":[158],"\"how":[159],"improve\"":[161],"\"what":[164],"is":[165],"correct.\"":[166],"Experimental":[167],"show":[169],"significantly":[172],"enhances":[173],"LLM":[174],"performance":[175],"across":[176],"ten":[177],"diverse":[178],"tasks":[179],"(avg.":[180],"49.6%":[182],"53.5%)":[184],"without":[185],"task-specific":[186],"fine-tuning.":[187],"Notably,":[188],"also":[194],"demonstrate":[195],"substantial":[196],"improvements":[197],"of":[201],"beyond":[203],"mere":[204],"accuracy,":[205],"suggesting":[206],"truly":[209],"teaches":[210],"self-improve":[213],"over":[214],"time.":[215]},"counts_by_year":[],"updated_date":"2026-09-30T09:27:27.532063","created_date":"2024-10-21T00:00:00"}