{"id":"https://openalex.org/W4415328503","doi":"https://doi.org/10.52202/085713-3573","title":"The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning","display_name":"The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4415328503","doi":"https://doi.org/10.52202/085713-3573"},"language":"en","primary_location":{"id":"doi:10.52202/085713-3573","doi":"https://doi.org/10.52202/085713-3573","is_oa":false,"landing_page_url":"https://doi.org/10.52202/085713-3573","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Advances in Neural Information Processing Systems 38","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2505.15134","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101442431","display_name":"Shivam Agarwal","orcid":"https://orcid.org/0000-0001-7221-1171","observed_orcids":["https://orcid.org/0000-0001-7221-1171"]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shivam Agarwal","raw_affiliation_strings":["University of Illinois at Urbana-Champaign"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Illinois at Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Zimin Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zimin Zhang","raw_affiliation_strings":["UIUC"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UIUC","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Lifan Yuan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Lifan Yuan","raw_affiliation_strings":["University of Illinois Urbana-Champaign"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jiawei Han","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jiawei Han","raw_affiliation_strings":["UIUC"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UIUC","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100740618","display_name":"Hao Ming Peng","orcid":"https://orcid.org/0000-0001-7422-630X","observed_orcids":["https://orcid.org/0000-0001-7422-630X"]},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hao Peng","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":3.793,"has_fulltext":true,"cited_by_count":4,"citation_normalized_percentile":{"value":0.94317607,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"118599","last_page":"118629"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8827999830245972,"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":0.8827999830245972,"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":0.04479999840259552,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.01720000058412552,"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/entropy-minimization","display_name":"entropy minimization","score":0.9990000128746033},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8349999785423279},{"id":"https://openalex.org/keywords/llm-reasoning","display_name":"LLM reasoning","score":0.7549999952316284},{"id":"https://openalex.org/keywords/self-consistency","display_name":"self-consistency","score":0.5299999713897705},{"id":"https://openalex.org/keywords/qwen","display_name":"Qwen","score":0.3610000014305115},{"id":"https://openalex.org/keywords/self-supervised-fine-tuning","display_name":"self-supervised fine-tuning","score":0.4699999988079071}],"concepts":[{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.6746000051498413},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.6363000273704529},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5636000037193298},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5315999984741211},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5239999890327454},{"id":"https://openalex.org/C9679016","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Principle of maximum entropy","level":2,"score":0.3912000060081482},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3513000011444092},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.33149999380111694},{"id":"https://openalex.org/C171752962","wikidata":"https://www.wikidata.org/wiki/Q255166","display_name":"Kullback\u2013Leibler divergence","level":2,"score":0.32739999890327454},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.32429999113082886},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3019999861717224},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3010999858379364},{"id":"https://openalex.org/C127233936","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Entropy maximization","level":3,"score":0.2989000082015991},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.296999990940094},{"id":"https://openalex.org/C2776330181","wikidata":"https://www.wikidata.org/wiki/Q18358244","display_name":"Maximization","level":2,"score":0.2624000012874603},{"id":"https://openalex.org/C107321475","wikidata":"https://www.wikidata.org/wiki/Q5374254","display_name":"Empirical risk minimization","level":2,"score":0.26190000772476196},{"id":"https://openalex.org/C2983703474","wikidata":"https://www.wikidata.org/wiki/Q17088227","display_name":"Probability estimation","level":2,"score":0.25600001215934753},{"id":"https://openalex.org/C2983203078","wikidata":"https://www.wikidata.org/wiki/Q255166","display_name":"Information gain","level":2,"score":0.25519999861717224}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.52202/085713-3573","doi":"https://doi.org/10.52202/085713-3573","is_oa":false,"landing_page_url":"https://doi.org/10.52202/085713-3573","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Advances in Neural Information Processing Systems 38","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2505.15134","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.15134","pdf_url":"https://arxiv.org/pdf/2505.15134","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2505.15134","doi":"https://doi.org/10.48550/arxiv.2505.15134","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2505.15134","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2505.15134","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.15134","pdf_url":"https://arxiv.org/pdf/2505.15134","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2063335192","display_name":"Molecule Maker Lab Institute (MMLI): An AI Institute for Molecular Discovery, Synthetic Strategy, and Manufacturing","funder_award_id":"2019897","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G3178773469","display_name":"III: Medium: Collaborative Research: Mining and Leveraging Knowledge Hypercubes for Complex Applications","funder_award_id":"1956151","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G3269309973","display_name":"HDR Institute: Geospatial Understanding through an Integrative Discovery Environment","funder_award_id":"2118329","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G3904442369","display_name":null,"funder_award_id":"HR0011-24-3-0325","funder_id":"https://openalex.org/F4320332180","funder_display_name":"Defense Advanced Research Projects Agency"},{"id":"https://openalex.org/G5781048298","display_name":"Category I: Bridging the Gap Between AI/ML Computing Demands and Today's Capabilities","funder_award_id":"2320345","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320317052","display_name":"Allen Institute for Artificial Intelligence","ror":"https://ror.org/05w520734"},{"id":"https://openalex.org/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"},{"id":"https://openalex.org/F4320332222","display_name":"University of Illinois at Urbana-Champaign","ror":"https://ror.org/047426m28"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415328503.pdf","grobid_xml":"https://content.openalex.org/works/W4415328503.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416529508","https://openalex.org/W4416049919","https://openalex.org/W7162079985","https://openalex.org/W7138946210","https://openalex.org/W7161816775","https://openalex.org/W4404396425","https://openalex.org/W4416383319","https://openalex.org/W4417086737","https://openalex.org/W4416131239","https://openalex.org/W4407424722"],"abstract_inverted_index":{"Entropy":[0],"minimization":[1,176],"(EM)":[2],"trains":[3],"the":[4,62,72,128,143],"model":[5],"to":[6,53,75,82,124],"concentrate":[7],"even":[8,183],"more":[9,150],"probability":[10],"mass":[11],"on":[12,36,57,116,142],"its":[13],"most":[14],"confident":[15],"outputs.":[16],"We":[17,43],"show":[18],"that":[19,113,160,169],"this":[20],"simple":[21],"objective":[22],"alone,":[23,177],"without":[24,85,95,178],"any":[25,86,96,179,184],"labeled":[26,97,118,180],"data,":[27,98],"can":[28,170],"substantially":[29],"improve":[30],"large":[31],"language":[32],"models'":[33],"(LLMs)":[34],"performance":[35,103,129],"challenging":[37,144],"math,":[38],"physics,":[39],"and":[40,111,138,154],"coding":[41],"tasks.":[42],"explore":[44],"three":[45],"approaches:":[46],"(1)":[47],"EM-FT":[48],"minimizes":[49],"token-level":[50],"entropy":[51,70,84,175],"similarly":[52],"instruction":[54],"finetuning,":[55],"but":[56],"unlabeled":[58],"outputs":[59],"drawn":[60],"from":[61],"model;":[63],"(2)":[64],"EM-RL:":[65],"reinforcement":[66],"learning":[67],"with":[68],"negative":[69],"as":[71,109],"only":[73],"reward":[74],"maximize;":[76],"(3)":[77],"EM-INF:":[78],"inference-time":[79],"logit":[80],"adjustment":[81],"reduce":[83],"training":[87],"data":[88,181],"or":[89,101,126,182],"parameter":[90,185],"updates.":[91,186],"On":[92],"Qwen-7B,":[93],"EM-RL,":[94],"achieves":[99],"comparable":[100],"better":[102],"than":[104,152],"strong":[105],"RL":[106],"baselines":[107],"such":[108],"GRPO":[110],"RLOO":[112],"are":[114],"trained":[115],"60K":[117],"examples.":[119],"Furthermore,":[120],"EM-INF":[121],"enables":[122],"Qwen-32B":[123],"match":[125],"exceed":[127],"of":[130],"proprietary":[131],"models":[132],"like":[133],"GPT-4o,":[134],"Claude":[135],"3":[136],"Opus,":[137],"Gemini":[139],"1.5":[140],"Pro":[141],"SciCode":[145],"benchmark,":[146],"while":[147],"being":[148],"3x":[149],"efficient":[151],"self-consistency":[153],"sequential":[155],"refinement.":[156],"Our":[157],"findings":[158],"reveal":[159],"many":[161],"pretrained":[162],"LLMs":[163],"possess":[164],"previously":[165],"underappreciated":[166],"reasoning":[167],"capabilities":[168],"be":[171],"effectively":[172],"elicited":[173],"through":[174]},"counts_by_year":[{"year":2026,"cited_by_count":3},{"year":2025,"cited_by_count":1}],"updated_date":"2026-10-21T09:56:26.102985","created_date":"2025-10-19T00:00:00"}