{"id":"https://openalex.org/W4402712382","doi":"https://doi.org/10.48550/arxiv.2408.16737","title":"Smaller, Weaker, Yet Better: Training LLM Reasoners via Compute-Optimal Sampling","display_name":"Smaller, Weaker, Yet Better: Training LLM Reasoners via Compute-Optimal Sampling","publication_year":2024,"publication_date":"2024-08-29","ids":{"openalex":"https://openalex.org/W4402712382","doi":"https://doi.org/10.48550/arxiv.2408.16737"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2408.16737","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.16737","pdf_url":"https://arxiv.org/pdf/2408.16737","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2408.16737","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5112428050","display_name":"Hritik Bansal","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Bansal, Hritik","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016112167","display_name":"Arian Hosseini","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hosseini, Arian","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109786478","display_name":"Rishabh Agarwal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Agarwal, Rishabh","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113394707","display_name":"Vinh Q. Tran","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tran, Vinh Q.","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5113394708","display_name":"Mehran Kazemi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kazemi, Mehran","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5112428050"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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/T13643","display_name":"Artificial Intelligence in Law","score":0.9502000212669373,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"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/T13643","display_name":"Artificial Intelligence in Law","score":0.9502000212669373,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6667491793632507},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4736422598361969},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.4734330177307129},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.35454264283180237},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.1480303704738617},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.09201601147651672}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6667491793632507},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4736422598361969},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.4734330177307129},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.35454264283180237},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.1480303704738617},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.09201601147651672},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2408.16737","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.16737","pdf_url":"https://arxiv.org/pdf/2408.16737","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},{"id":"doi:10.48550/arxiv.2408.16737","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2408.16737","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2408.16737","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.16737","pdf_url":"https://arxiv.org/pdf/2408.16737","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W4391375266","https://openalex.org/W1979597421","https://openalex.org/W2007980826","https://openalex.org/W2061531152","https://openalex.org/W3002753104","https://openalex.org/W2077600819","https://openalex.org/W2142036596","https://openalex.org/W2072657027","https://openalex.org/W2600246793"],"abstract_inverted_index":{"Training":[0],"on":[1,104,139,146,168],"high-quality":[2],"synthetic":[3,47,172],"data":[4,48,68,83,105,141,148,173],"from":[5,84,106],"strong":[6],"language":[7],"models":[8,86,110,137,170],"(LMs)":[9],"is":[10,29],"a":[11,32,50,58,118,124,130],"common":[12],"strategy":[13,28],"to":[14,129],"improve":[15],"the":[16,43,66,82,163,180],"reasoning":[17,128],"performance":[18],"of":[19,155,166],"LMs.":[20],"In":[21],"this":[22,27],"work,":[23],"we":[24,41],"revisit":[25],"whether":[26],"compute-optimal":[30,181],"under":[31],"fixed":[33],"inference":[34],"budget":[35],"(e.g.,":[36],"FLOPs).":[37],"To":[38],"do":[39],"so,":[40],"investigate":[42],"trade-offs":[44],"between":[45],"generating":[46],"using":[49],"stronger":[51,131],"but":[52,60,93],"more":[53],"expensive":[54],"(SE)":[55],"model":[56],"versus":[57],"weaker":[59,125],"cheaper":[61],"(WC)":[62],"model.":[63],"We":[64,100],"evaluate":[65],"generated":[67],"across":[69,149],"three":[70],"key":[71],"metrics:":[72],"coverage,":[73],"diversity,":[74,92],"and":[75,79,91,108,117,152,157],"false":[76,97],"positive":[77,98],"rate,":[78],"show":[80],"that":[81,136,176],"WC":[85,109,156,177],"may":[87,178],"have":[88],"higher":[89,96],"coverage":[90],"also":[94],"exhibit":[95],"rates.":[99],"then":[101],"finetune":[102],"LMs":[103],"SE":[107,158,169],"in":[111],"different":[112],"settings:":[113],"knowledge":[114],"distillation,":[115],"self-improvement,":[116],"novel":[119],"weak-to-strong":[120],"improvement":[121],"setup":[122],"where":[123],"LM":[126,186],"teaches":[127],"LM.":[132],"Our":[133],"findings":[134],"reveal":[135],"finetuned":[138],"WC-generated":[140],"consistently":[142],"outperform":[143],"those":[144],"trained":[145],"SE-generated":[147],"multiple":[150,153],"benchmarks":[151],"choices":[154],"models.":[159],"These":[160],"results":[161],"challenge":[162],"prevailing":[164],"practice":[165],"relying":[167],"for":[171,183],"generation,":[174],"suggesting":[175],"be":[179],"approach":[182],"training":[184],"advanced":[185],"reasoners.":[187]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-03-10T16:38:18.471706","created_date":"2024-09-22T00:00:00"}