{"id":"https://openalex.org/W7137849016","doi":"https://doi.org/10.48550/arxiv.2603.13305","title":"Evidence-based Distributional Alignment for Large Language Models","display_name":"Evidence-based Distributional Alignment for Large Language Models","publication_year":2026,"publication_date":"2026-03-03","ids":{"openalex":"https://openalex.org/W7137849016","doi":"https://doi.org/10.48550/arxiv.2603.13305"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.13305","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13305","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":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.2603.13305","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5105938104","display_name":"Viet-Thanh Pham","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pham, Viet-Thanh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129753192","display_name":"Lizhen Qu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qu, Lizhen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129724395","display_name":"Zhuang Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhuang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5129647508","display_name":"Gholamreza Haffari","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haffari, Gholamreza","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/T13910","display_name":"Computational and Text Analysis Methods","score":0.48820000886917114,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"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/T13910","display_name":"Computational and Text Analysis Methods","score":0.48820000886917114,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.1615999937057495,"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/T13274","display_name":"Expert finding and Q&A systems","score":0.02019999921321869,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/robustness","display_name":"Robustness (evolution)","score":0.7336000204086304},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.5455999970436096},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5230000019073486},{"id":"https://openalex.org/keywords/population","display_name":"Population","score":0.4742000102996826},{"id":"https://openalex.org/keywords/divergence","display_name":"Divergence (linguistics)","score":0.4632999897003174},{"id":"https://openalex.org/keywords/value","display_name":"Value (mathematics)","score":0.3824999928474426},{"id":"https://openalex.org/keywords/distribution","display_name":"Distribution (mathematics)","score":0.3671000003814697},{"id":"https://openalex.org/keywords/kullback\u2013leibler-divergence","display_name":"Kullback\u2013Leibler divergence","score":0.3546999990940094}],"concepts":[{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.7336000204086304},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6004999876022339},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.5455999970436096},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5230000019073486},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48489999771118164},{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.4742000102996826},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.4632999897003174},{"id":"https://openalex.org/C2776291640","wikidata":"https://www.wikidata.org/wiki/Q2912517","display_name":"Value (mathematics)","level":2,"score":0.3824999928474426},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3707999885082245},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.3671000003814697},{"id":"https://openalex.org/C171752962","wikidata":"https://www.wikidata.org/wiki/Q255166","display_name":"Kullback\u2013Leibler divergence","level":2,"score":0.3546999990940094},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.3418999910354614},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.31439998745918274},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3138999938964844},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.31200000643730164},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.30880001187324524},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2831000089645386},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.2815000116825104},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.2777000069618225},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.27489998936653137}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.13305","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13305","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":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.2603.13305","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13305","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Distributional":[0],"alignment":[1,78],"enables":[2],"large":[3],"language":[4],"models":[5],"(LLMs)":[6],"to":[7,61,182,191],"predict":[8],"how":[9],"a":[10,24,96,100,115,130,138],"target":[11,97],"population":[12],"distributes":[13],"its":[14],"responses":[15],"across":[16],"answer":[17,112,127],"options,":[18],"rather":[19],"than":[20],"collapsing":[21],"disagreement":[22],"into":[23],"single":[25],"consensus":[26],"answer.":[27],"However,":[28],"existing":[29],"LLM-based":[30,87],"distribution":[31,88,128],"prediction":[32],"is":[33],"often":[34],"unstable":[35],"and":[36,40,59,63,66,84,92,99,110,123,159,165,168,178],"degrades":[37],"under":[38,90],"cultural":[39,93,161],"domain":[41,91],"shift.":[42,94],"Token":[43],"score-based":[44],"estimates":[45,56],"can":[46],"change":[47],"with":[48,185],"minor":[49],"option":[50],"wording":[51],"or":[52],"formatting,":[53],"response":[54],"sampling-based":[55],"are":[57,70],"expensive":[58],"sensitive":[60],"prompts":[62],"decoding":[64],"settings,":[65],"directly":[67],"generated":[68],"distributions":[69,180],"frequently":[71],"miscalibrated.":[72],"We":[73,133],"propose":[74],"Evi-DA,":[75],"an":[76],"evidence-based":[77],"technique":[79],"that":[80,147],"improves":[81],"the":[82,125,135],"fidelity":[83],"robustness":[85],"of":[86,189],"estimation":[89],"Given":[95],"country":[98],"multiple-choice":[101],"question,":[102],"Evi-DA":[103,172],"retrieves":[104],"related":[105],"World":[106],"Values":[107],"Survey":[108],"items":[109],"their":[111],"distributions,":[113,155],"predicts":[114],"coarse":[116],"Welzel":[117],"value":[118,151],"signature":[119],"for":[120],"each":[121],"option,":[122],"infers":[124],"country-conditioned":[126],"in":[129],"structured":[131,157],"format.":[132],"train":[134],"LLMs":[136],"using":[137],"two-stage":[139],"pipeline,":[140],"where":[141],"reinforcement":[142],"learning":[143],"optimizes":[144],"survey-derived":[145],"rewards":[146],"encourage":[148],"accurate":[149],"intermediate":[150],"predictions,":[152],"faithful":[153],"final":[154],"well-formed":[156],"outputs,":[158],"reduced":[160],"bias.":[162],"Across":[163],"in-domain":[164],"out-of-domain":[166],"benchmarks":[167],"multiple":[169],"open-source":[170],"backbones,":[171],"reduces":[173],"Jensen-Shannon":[174],"divergence":[175],"between":[176],"predicted":[177],"gold":[179],"relative":[181,187],"strong":[183],"baselines,":[184],"average":[186],"improvements":[188],"up":[190],"44%.":[192]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-18T00:00:00"}