{"id":"https://openalex.org/W4407368670","doi":"https://doi.org/10.48550/arxiv.2502.05234","title":"Optimizing Temperature for Language Models with Multi-Sample Inference","display_name":"Optimizing Temperature for Language Models with Multi-Sample Inference","publication_year":2025,"publication_date":"2025-02-07","ids":{"openalex":"https://openalex.org/W4407368670","doi":"https://doi.org/10.48550/arxiv.2502.05234"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2502.05234","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.05234","pdf_url":"https://arxiv.org/pdf/2502.05234","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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2502.05234","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101541718","display_name":"Weihua Du","orcid":"https://orcid.org/0000-0002-8856-0277","observed_orcids":["https://orcid.org/0000-0002-8856-0277"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Weihua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101421479","display_name":"Yiming Yang","orcid":"https://orcid.org/0000-0002-9755-2711","observed_orcids":["https://orcid.org/0000-0002-9755-2711"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Yiming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5019030424","display_name":"Sean Welleck","orcid":"https://orcid.org/0009-0003-1744-0096","observed_orcids":["https://orcid.org/0009-0003-1744-0096"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Welleck, Sean","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":33,"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":0.8862000107765198,"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":0.8862000107765198,"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":0.0835999995470047,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.011900000274181366,"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/temperature-scaling","display_name":"temperature scaling","score":0.8100000023841858},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.875},{"id":"https://openalex.org/keywords/majority-voting","display_name":"majority voting","score":0.718999981880188},{"id":"https://openalex.org/keywords/best-of-n-sampling","display_name":"best-of-N sampling","score":0.9929999709129333},{"id":"https://openalex.org/keywords/entropy-measures","display_name":"entropy measures","score":0.9459999799728394},{"id":"https://openalex.org/keywords/stochastic-modeling","display_name":"stochastic modeling","score":0.9070000052452087}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7447341084480286},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.5883418321609497},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4489459693431854},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.40139448642730713},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.3841993808746338},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.35025644302368164},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3478736877441406},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.21498501300811768},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.12442976236343384},{"id":"https://openalex.org/C97355855","wikidata":"https://www.wikidata.org/wiki/Q11473","display_name":"Thermodynamics","level":1,"score":0.07872983813285828},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.07685256004333496}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2502.05234","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.05234","pdf_url":"https://arxiv.org/pdf/2502.05234","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.2502.05234","doi":"https://doi.org/10.48550/arxiv.2502.05234","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2502.05234","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:2502.05234","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.05234","pdf_url":"https://arxiv.org/pdf/2502.05234","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":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4407368670.pdf","grobid_xml":"https://content.openalex.org/works/W4407368670.grobid-xml"},"referenced_works_count":31,"referenced_works":["https://openalex.org/W2918914336","https://openalex.org/W2938704169","https://openalex.org/W3134642945","https://openalex.org/W3177813494","https://openalex.org/W3195140297","https://openalex.org/W4221161695","https://openalex.org/W4383473935","https://openalex.org/W4385569780","https://openalex.org/W4386185625","https://openalex.org/W4387561528","https://openalex.org/W4387725649","https://openalex.org/W4391272793","https://openalex.org/W4391631327","https://openalex.org/W4393119187","https://openalex.org/W4397028247","https://openalex.org/W4400024829","https://openalex.org/W4400343354","https://openalex.org/W4400703259","https://openalex.org/W4401306886","https://openalex.org/W4401307656","https://openalex.org/W4402670055","https://openalex.org/W4403524626","https://openalex.org/W4403796339","https://openalex.org/W4403928588","https://openalex.org/W4404451370","https://openalex.org/W4404781852","https://openalex.org/W4404782267","https://openalex.org/W4405766390","https://openalex.org/W4411120404","https://openalex.org/W4411638692","https://openalex.org/W4415797700"],"related_works":["https://openalex.org/W4413167964","https://openalex.org/W4393119187","https://openalex.org/W4404781852","https://openalex.org/W7147350604","https://openalex.org/W7128812669","https://openalex.org/W4416369109","https://openalex.org/W4414814559","https://openalex.org/W7221809747","https://openalex.org/W7213575038","https://openalex.org/W7147711460"],"abstract_inverted_index":{"Multi-sample":[0],"aggregation":[1,80],"strategies,":[2,81],"such":[3],"as":[4],"majority":[5],"voting":[6],"and":[7,60,109,146],"best-of-N":[8],"sampling,":[9],"are":[10,57],"widely":[11],"used":[12],"in":[13,29,96,101],"contemporary":[14],"large":[15],"language":[16],"models":[17],"(LLMs)":[18],"to":[19,62,135],"enhance":[20,136],"predictive":[21,110],"accuracy":[22],"across":[23],"various":[24],"tasks.":[25],"A":[26],"key":[27],"challenge":[28,68],"this":[30],"process":[31,133],"is":[32],"temperature":[33,48,74,121,145],"selection,":[34],"which":[35,56,123],"significantly":[36],"impacts":[37],"model":[38,102,107,134,147],"performance.":[39,148],"Existing":[40],"approaches":[41],"either":[42],"rely":[43],"on":[44,84],"a":[45,90,115,131],"fixed":[46],"default":[47],"or":[49],"require":[50],"labeled":[51],"validation":[52,86],"data":[53],"for":[54,75,119],"tuning,":[55],"often":[58],"scarce":[59],"difficult":[61],"obtain.":[63],"This":[64],"paper":[65],"addresses":[66],"the":[67,72,142],"of":[69,93],"automatically":[70],"identifying":[71],"(near)-optimal":[73],"different":[76],"LLMs":[77],"using":[78],"multi-sample":[79],"without":[82],"relying":[83],"task-specific":[85],"data.":[87],"We":[88],"provide":[89],"comprehensive":[91],"analysis":[92],"temperature's":[94],"role":[95],"performance":[97],"optimization,":[98,122],"considering":[99],"variations":[100],"architectures,":[103],"datasets,":[104],"task":[105],"types,":[106],"sizes,":[108],"accuracy.":[111],"Furthermore,":[112],"we":[113,129],"propose":[114],"novel":[116],"entropy-based":[117],"metric":[118],"automated":[120],"consistently":[124],"outperforms":[125],"fixed-temperature":[126],"baselines.":[127],"Additionally,":[128],"incorporate":[130],"stochastic":[132],"interpretability,":[137],"offering":[138],"deeper":[139],"insights":[140],"into":[141],"relationship":[143],"between":[144]},"counts_by_year":[{"year":2026,"cited_by_count":18},{"year":2025,"cited_by_count":15}],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2025-02-12T00:00:00"}