{"id":"https://openalex.org/W4415057368","doi":"https://doi.org/10.48550/arxiv.2509.06941","title":"Outcome-based Exploration for LLM Reasoning","display_name":"Outcome-based Exploration for LLM Reasoning","publication_year":2025,"publication_date":"2025-09-08","ids":{"openalex":"https://openalex.org/W4415057368","doi":"https://doi.org/10.48550/arxiv.2509.06941"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2509.06941","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.06941","pdf_url":"https://arxiv.org/pdf/2509.06941","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/2509.06941","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111605548","display_name":"Yu Song","orcid":"https://orcid.org/0009-0006-1964-5281","observed_orcids":["https://orcid.org/0009-0006-1964-5281"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Yuda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5051933322","display_name":"Julia Kempe","orcid":"https://orcid.org/0009-0006-1984-0060","observed_orcids":["https://orcid.org/0009-0006-1984-0060","https://orcid.org/0000-0003-0448-5941"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kempe, Julia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5006533777","display_name":"R\u00e9mi Munos","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Munos, Remi","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":false,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.5264000296592712,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.5264000296592712,"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/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.1542000025510788,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.12039999663829803,"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/large-language-models","display_name":"large language models","score":0.7789999842643738},{"id":"https://openalex.org/keywords/llm-reasoning","display_name":"LLM reasoning","score":0.8169999718666077},{"id":"https://openalex.org/keywords/reinforcement-learning-post-training","display_name":"reinforcement learning post-training","score":0.7889999747276306},{"id":"https://openalex.org/keywords/generational-diversity","display_name":"generational diversity","score":0.7179999947547913},{"id":"https://openalex.org/keywords/umbilical-cord-blood","display_name":"umbilical cord blood","score":0.6899999976158142},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.39399999380111694}],"concepts":[{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.7745000123977661},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6797000169754028},{"id":"https://openalex.org/C2781316041","wikidata":"https://www.wikidata.org/wiki/Q1230584","display_name":"Diversity (politics)","level":2,"score":0.6546000242233276},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6186000108718872},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5794000029563904},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5746999979019165},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5393000245094299},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5016000270843506},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.43869999051094055},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.4163999855518341},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.37540000677108765},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3718000054359436},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.36649999022483826},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.34139999747276306},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.32919999957084656},{"id":"https://openalex.org/C91306197","wikidata":"https://www.wikidata.org/wiki/Q45767","display_name":"Competition (biology)","level":2,"score":0.3190999925136566},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.2867000102996826},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.2827000021934509},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.274399995803833},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.272599995136261},{"id":"https://openalex.org/C50335755","wikidata":"https://www.wikidata.org/wiki/Q483247","display_name":"Phenomenon","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C190812933","wikidata":"https://www.wikidata.org/wiki/Q28923","display_name":"Chart","level":2,"score":0.2581000030040741},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.2547999918460846}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2509.06941","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.06941","pdf_url":"https://arxiv.org/pdf/2509.06941","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.2509.06941","doi":"https://doi.org/10.48550/arxiv.2509.06941","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2509.06941","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:2509.06941","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.06941","pdf_url":"https://arxiv.org/pdf/2509.06941","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416611553","https://openalex.org/W7127425756","https://openalex.org/W4416766928","https://openalex.org/W4417144671","https://openalex.org/W7124227831","https://openalex.org/W7220664336","https://openalex.org/W7127960403","https://openalex.org/W7083817313","https://openalex.org/W7160888893","https://openalex.org/W7164152851"],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"(RL)":[2],"has":[3],"emerged":[4],"as":[5,65],"a":[6,39,66,95,123,203,213],"powerful":[7],"method":[8],"for":[9,25,54,227],"improving":[10],"the":[11,26,29,80,85,112,115,192,197,224],"reasoning":[12,119,221],"abilities":[13],"of":[14,28,97,114,126,199,206],"large":[15],"language":[16],"models":[17,180],"(LLMs).":[18],"Outcome-based":[19],"RL,":[20],"which":[21,137,152,163],"rewards":[22],"policies":[23],"solely":[24],"correctness":[27],"final":[30,143],"answer,":[31],"yields":[32],"substantial":[33],"accuracy":[34,186],"gains":[35],"but":[36],"also":[37],"induces":[38],"systematic":[40],"loss":[41],"in":[42],"generation":[43],"diversity.":[44,170],"This":[45],"collapse":[46],"undermines":[47],"real-world":[48],"performance,":[49],"where":[50,100],"diversity":[51,77,98,102,189,225],"is":[52],"critical":[53],"test-time":[55,169],"scaling.":[56],"We":[57,145],"analyze":[58],"this":[59],"phenomenon":[60],"by":[61,130],"viewing":[62],"RL":[63,73,217],"post-training":[64],"sampling":[67],"process":[68],"and":[69,110,160,178],"show":[70],"that,":[71],"strikingly,":[72],"can":[74],"reduce":[75],"effective":[76],"even":[78],"on":[79,103,172],"training":[81],"set":[82,125],"relative":[83],"to":[84,107,142,167],"base":[86],"model.":[87],"Our":[88],"study":[89],"highlights":[90],"two":[91,147],"central":[92],"findings:":[93],"(i)":[94],"transfer":[96],"degradation,":[99],"reduced":[101],"solved":[104],"problems":[105],"propagates":[106],"unsolved":[108],"ones,":[109],"(ii)":[111],"tractability":[113],"outcome":[116],"space,":[117],"since":[118],"tasks":[120],"admit":[121],"only":[122],"limited":[124],"distinct":[127],"answers.":[128],"Motivated":[129],"these":[131,210],"insights,":[132],"we":[133,195],"propose":[134],"outcome-based":[135,200,207],"exploration,":[136,151,162],"assigns":[138],"exploration":[139,201],"bonuses":[140],"according":[141],"outcomes.":[144],"introduce":[146],"complementary":[148],"algorithms:":[149],"historical":[150],"encourages":[153],"rarely":[154],"observed":[155],"answers":[156],"via":[157],"UCB-style":[158],"bonuses,":[159],"batch":[161],"penalizes":[164],"within-batch":[165],"repetition":[166],"promote":[168],"Experiments":[171],"standard":[173],"competition":[174],"math":[175],"with":[176],"Llama":[177],"Qwen":[179],"demonstrate":[181],"that":[182,219],"both":[183],"methods":[184,218],"improve":[185],"while":[187],"mitigating":[188],"collapse.":[190],"On":[191],"theoretical":[193],"side,":[194],"formalize":[196],"benefit":[198],"through":[202],"new":[204],"model":[205],"bandits.":[208],"Together,":[209],"contributions":[211],"chart":[212],"practical":[214],"path":[215],"toward":[216],"enhance":[220],"without":[222],"sacrificing":[223],"essential":[226],"scalable":[228],"deployment.":[229]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-09-24T07:57:37.273135","created_date":"2025-10-11T00:00:00"}