{"id":"https://openalex.org/W4417077993","doi":"https://doi.org/10.48550/arxiv.2509.15194","title":"Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation","display_name":"Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation","publication_year":2025,"publication_date":"2025-09-18","ids":{"openalex":"https://openalex.org/W4417077993","doi":"https://doi.org/10.48550/arxiv.2509.15194"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2509.15194","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.15194","pdf_url":"https://arxiv.org/pdf/2509.15194","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,"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"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2509.15194","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5028930388","display_name":"Yujun Zhou","orcid":"https://orcid.org/0000-0001-6571-3691"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Yujun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Liang, Zhenwen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Zhenwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100674213","display_name":"Haolin Liu","orcid":"https://orcid.org/0000-0003-3192-6378"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Haolin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072237514","display_name":"Wenhao Yu","orcid":"https://orcid.org/0000-0003-2991-5362"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Wenhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005536479","display_name":"Kishan Panaganti","orcid":"https://orcid.org/0000-0001-9746-698X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Panaganti, Kishan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016877422","display_name":"Linfeng Song","orcid":"https://orcid.org/0000-0002-3502-3574"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Linfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113226487","display_name":"Dian Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Dian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000755750","display_name":"Xiangliang Zhang","orcid":"https://orcid.org/0000-0002-3574-5665"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xiangliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071533156","display_name":"Haitao Mi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mi, Haitao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5034476404","display_name":"Dong Yu","orcid":"https://orcid.org/0000-0003-0520-6844"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Dong","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":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":0.2985999882221222,"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":0.2985999882221222,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.11869999766349792,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06930000334978104,"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/novelty","display_name":"Novelty","score":0.7745000123977661},{"id":"https://openalex.org/keywords/variation","display_name":"Variation (astronomy)","score":0.6198999881744385},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.5612000226974487},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.551800012588501},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.5407999753952026},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.49869999289512634},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4648999869823456}],"concepts":[{"id":"https://openalex.org/C2778738651","wikidata":"https://www.wikidata.org/wiki/Q16546687","display_name":"Novelty","level":2,"score":0.7745000123977661},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6593000292778015},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.6198999881744385},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.600600004196167},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5612000226974487},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.551800012588501},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.5407999753952026},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.49869999289512634},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4648999869823456},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.42500001192092896},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.3720000088214874},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.36000001430511475},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.3328000009059906},{"id":"https://openalex.org/C9679016","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Principle of maximum entropy","level":2,"score":0.31150001287460327},{"id":"https://openalex.org/C2781316041","wikidata":"https://www.wikidata.org/wiki/Q1230584","display_name":"Diversity (politics)","level":2,"score":0.3086000084877014},{"id":"https://openalex.org/C2778924833","wikidata":"https://www.wikidata.org/wiki/Q7064603","display_name":"Novelty detection","level":3,"score":0.288100004196167},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.26170000433921814}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2509.15194","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.15194","pdf_url":"https://arxiv.org/pdf/2509.15194","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,"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.2509.15194","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2509.15194","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2509.15194","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.15194","pdf_url":"https://arxiv.org/pdf/2509.15194","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,"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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,18,46],"(LLMs)":[3],"are":[4],"increasingly":[5],"trained":[6],"with":[7,78],"reinforcement":[8],"learning":[9],"from":[10,107,147,154],"verifiable":[11],"rewards":[12],"(RLVR),":[13],"yet":[14],"real-world":[15],"deployment":[16],"demands":[17],"that":[19,55,70,96,131],"can":[20],"self-improve":[21],"without":[22],"labels":[23],"or":[24,38],"external":[25],"judges.":[26],"Existing":[27],"self-improvement":[28],"approaches":[29],"primarily":[30],"rely":[31],"on":[32,140],"self-confirmation":[33],"signals":[34],"(e.g.,":[35],"confidence,":[36],"entropy,":[37],"consistency)":[39],"to":[40,150,156,173],"generate":[41],"rewards.":[42],"This":[43,112],"reliance":[44],"drives":[45],"toward":[47],"over-confident,":[48],"majority-favored":[49],"solutions,":[50],"causing":[51],"an":[52,87],"entropy":[53],"collapse":[54,164],"degrades":[56],"pass@n":[57],"and":[58,152,178],"reasoning":[59,105,172],"complexity.":[60],"To":[61],"address":[62],"this,":[63],"we":[64],"propose":[65],"EVOL-RL,":[66],"a":[67,93],"label-free":[68,141],"framework":[69],"mirrors":[71,117],"the":[72,83,118,135],"evolutionary":[73],"principle":[74],"of":[75],"balancing":[76],"selection":[77,121],"variation.":[79],"Concretely,":[80],"EVOL-RL":[81,132,158],"retains":[82],"majority-voted":[84],"answer":[85],"as":[86],"anchor":[88],"for":[89],"stability,":[90],"but":[91,165],"adds":[92],"novelty-aware":[94],"reward":[95],"scores":[97],"each":[98],"sampled":[99],"solution":[100],"by":[101],"how":[102],"different":[103],"its":[104],"is":[106,182],"other":[108],"concurrently":[109],"generated":[110],"responses.":[111],"majority-for-stability":[113],"+":[114],"novelty-for-exploration":[115],"rule":[116],"variation-selection":[119],"principle:":[120],"prevents":[122,126,161],"drift,":[123],"while":[124],"novelty":[125],"collapse.":[127],"Evaluation":[128],"results":[129],"show":[130],"consistently":[133],"outperforms":[134],"majority-only":[136],"baseline;":[137],"e.g.,":[138,176],"training":[139],"AIME24":[142],"lifts":[143],"Qwen3-4B-Base":[144],"AIME25":[145],"pass@1":[146],"baseline's":[148],"4.6%":[149],"16.4%,":[151],"pass@16":[153],"18.5%":[155],"37.9%.":[157],"not":[159],"only":[160],"in-domain":[162],"diversity":[163],"also":[166],"improves":[167],"out-of-domain":[168],"generalization":[169],"(from":[170],"math":[171],"broader":[174],"tasks,":[175],"MMLU-Pro":[177],"BBEH).":[179],"The":[180],"code":[181],"available":[183],"at:":[184],"https://github.com/YujunZhou/EVOL-RL.":[185]},"counts_by_year":[],"updated_date":"2026-08-26T07:47:46.906454","created_date":"2025-10-10T00:00:00"}