{"id":"https://openalex.org/W7162670860","doi":"https://doi.org/10.48550/arxiv.2605.27747","title":"Soft Specialists: $\u03b1$-R\u00e9nyi Ensembles for Uncertainty-Aware LLM Post-Training","display_name":"Soft Specialists: $\u03b1$-R\u00e9nyi Ensembles for Uncertainty-Aware LLM Post-Training","publication_year":2026,"publication_date":"2026-05-26","ids":{"openalex":"https://openalex.org/W7162670860","doi":"https://doi.org/10.48550/arxiv.2605.27747"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.27747","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27747","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":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.27747","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114337203","display_name":"Paula Cordero-Encinar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cordero-Encinar, Paula","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5095335088","display_name":"Georgy Tyukin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tyukin, Georgy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5090813467","display_name":"A. Duncan","orcid":"https://orcid.org/0000-0001-5762-164X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Duncan, Andrew B.","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":0.34040001034736633,"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":0.34040001034736633,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.08179999887943268,"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"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.07429999858140945,"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/stability","display_name":"Stability (learning theory)","score":0.621999979019165},{"id":"https://openalex.org/keywords/ensemble-learning","display_name":"Ensemble learning","score":0.5974000096321106},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5674999952316284},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5138999819755554},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.5041000247001648},{"id":"https://openalex.org/keywords/bayes-theorem","display_name":"Bayes' theorem","score":0.4853000044822693},{"id":"https://openalex.org/keywords/posterior-probability","display_name":"Posterior probability","score":0.4120999872684479},{"id":"https://openalex.org/keywords/uncertainty-quantification","display_name":"Uncertainty quantification","score":0.3928000032901764}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.626800000667572},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.621999979019165},{"id":"https://openalex.org/C45942800","wikidata":"https://www.wikidata.org/wiki/Q245652","display_name":"Ensemble learning","level":2,"score":0.5974000096321106},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5965999960899353},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5674999952316284},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5411999821662903},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5138999819755554},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.5041000247001648},{"id":"https://openalex.org/C207201462","wikidata":"https://www.wikidata.org/wiki/Q182505","display_name":"Bayes' theorem","level":3,"score":0.4853000044822693},{"id":"https://openalex.org/C57830394","wikidata":"https://www.wikidata.org/wiki/Q278079","display_name":"Posterior probability","level":3,"score":0.4120999872684479},{"id":"https://openalex.org/C32230216","wikidata":"https://www.wikidata.org/wiki/Q7882499","display_name":"Uncertainty quantification","level":2,"score":0.3928000032901764},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.3871999979019165},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.3483000099658966},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3467000126838684},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.310699999332428},{"id":"https://openalex.org/C119898033","wikidata":"https://www.wikidata.org/wiki/Q3433888","display_name":"Ensemble forecasting","level":2,"score":0.30300000309944153},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.2825999855995178},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.2815999984741211},{"id":"https://openalex.org/C155846161","wikidata":"https://www.wikidata.org/wiki/Q1143367","display_name":"Graphical model","level":2,"score":0.27129998803138733},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.27000001072883606},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.26080000400543213},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2547999918460846},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.2540000081062317},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.27747","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27747","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":"doi:10.48550/arxiv.2605.27747","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27747","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":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.5310143828392029,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Existing":[0],"training":[1,141,153],"approaches":[2],"for":[3,55,143],"large":[4,15],"language":[5],"models":[6,88],"learn":[7],"a":[8,29,43,133,139],"single":[9],"set":[10],"of":[11,17,47,91,128],"parameters,":[12,60],"based":[13],"on":[14],"volumes":[16],"data,":[18],"which":[19],"is":[20,33],"typically":[21],"heterogeneous,":[22],"conflicting":[23,37,113],"and":[24,39,78,147,165],"often":[25],"outright":[26],"contradictory.":[27],"As":[28],"result,":[30],"the":[31],"model":[32,101,163],"forced":[34],"to":[35,65,122,132,155],"compress":[36],"goals,":[38],"inherent":[40],"uncertainties":[41],"into":[42],"single,":[44],"averaged":[45],"pattern":[46],"behaviour.":[48],"We":[49,94,118],"propose":[50],"an":[51,62,126],"$\u03b1$-R\u00e9nyi":[52],"variational":[53,71,76],"framework":[54,121],"learning":[56,125],"distributions":[57],"over":[58],"post-training":[59],"offering":[61],"uncertainty-aware":[63],"alternative":[64],"deep":[66],"ensemble":[67,127,160],"approaches.":[68],"The":[69],"resulting":[70],"objective":[72],"interpolates":[73],"between":[74,84],"classical":[75],"Bayes":[77],"predictively":[79],"oriented":[80],"posterior":[81,106],"learning,":[82],"balancing":[83],"globally":[85],"plausible":[86],"individual":[87],"against":[89],"systems":[90],"complementary":[92],"specialists.":[93],"identify":[95],"local":[96],"stability":[97],"criteria,":[98],"demonstrating":[99],"how":[100],"misspecification":[102],"can":[103],"make":[104],"non-degenerate":[105],"spread":[107],"locally":[108],"favourable,":[109],"manifesting":[110],"contradictory":[111],"or":[112],"data":[114],"as":[115],"epistemic":[116],"uncertainty.":[117],"apply":[119],"our":[120],"LLM":[123],"post-training,":[124],"LoRA":[129],"adapters":[130],"attached":[131],"shared,":[134],"frozen":[135],"base":[136],"model,":[137],"providing":[138,166],"scalable":[140],"procedure":[142],"both":[144],"supervised":[145],"fine-tuning":[146],"preference":[148],"optimisation.":[149],"Our":[150],"approach":[151],"enables":[152],"examples":[154],"be":[156],"softly":[157],"routed":[158],"across":[159,170],"members,":[161],"promoting":[162],"specialisation":[164],"actionable":[167],"uncertainty":[168],"estimates":[169],"different":[171],"tasks.":[172]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-29T00:00:00"}