{"id":"https://openalex.org/W7220719053","doi":"https://doi.org/10.48550/arxiv.2610.05450","title":"Groupwise Distortion Guarantees for Preference-Based Alignment","display_name":"Groupwise Distortion Guarantees for Preference-Based Alignment","publication_year":2026,"publication_date":"2026-10-04","ids":{"openalex":"https://openalex.org/W7220719053","doi":"https://doi.org/10.48550/arxiv.2610.05450"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.05450","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05450","pdf_url":"https://arxiv.org/pdf/2610.05450","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.05450","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5157724017","display_name":"Jacob Brodkey","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Brodkey, Jacob","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5157673937","display_name":"Roberto Tamez","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tamez, Roberto","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5057693522","display_name":"Aaron Roth","orcid":"https://orcid.org/0000-0002-0586-0515","observed_orcids":["https://orcid.org/0000-0002-0586-0515"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Roth, Aaron","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I79576946"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10991","display_name":"Game Theory and Voting Systems","score":0.8852999806404114,"subfield":{"id":"https://openalex.org/subfields/2002","display_name":"Economics and Econometrics"},"field":{"id":"https://openalex.org/fields/20","display_name":"Economics, Econometrics and Finance"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10991","display_name":"Game Theory and Voting Systems","score":0.8852999806404114,"subfield":{"id":"https://openalex.org/subfields/2002","display_name":"Economics and Econometrics"},"field":{"id":"https://openalex.org/fields/20","display_name":"Economics, Econometrics and Finance"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12072","display_name":"Machine Learning and Algorithms","score":0.02590000070631504,"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":0.01940000057220459,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/preference-based-alignment","display_name":"preference-based alignment","score":0.953000009059906},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.8080000281333923},{"id":"https://openalex.org/keywords/distortion-bounds","display_name":"distortion bounds","score":0.5529999732971191},{"id":"https://openalex.org/keywords/social-welfare-welfare-economics","display_name":"social welfare (welfare economics)","score":0.8399999737739563},{"id":"https://openalex.org/keywords/bradley-terry-model","display_name":"Bradley-Terry model","score":0.6700000166893005},{"id":"https://openalex.org/keywords/pairwise-comparison","display_name":"pairwise comparison","score":0.7599999904632568},{"id":"https://openalex.org/keywords/sample-complexity","display_name":"sample complexity","score":0.7210000157356262}],"concepts":[{"id":"https://openalex.org/C126780896","wikidata":"https://www.wikidata.org/wiki/Q899871","display_name":"Distortion (music)","level":4,"score":0.7796000242233276},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.7657999992370605},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.5346999764442444},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4805000126361847},{"id":"https://openalex.org/C181789720","wikidata":"https://www.wikidata.org/wiki/Q4812191","display_name":"Asymptotically optimal algorithm","level":2,"score":0.4740000069141388},{"id":"https://openalex.org/C100243477","wikidata":"https://www.wikidata.org/wiki/Q12002092","display_name":"Welfare","level":2,"score":0.4717999994754791},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.4717000126838684},{"id":"https://openalex.org/C2781311116","wikidata":"https://www.wikidata.org/wiki/Q83306","display_name":"Group (periodic table)","level":2,"score":0.4562000036239624},{"id":"https://openalex.org/C189352744","wikidata":"https://www.wikidata.org/wiki/Q1971461","display_name":"Social choice theory","level":2,"score":0.45100000500679016},{"id":"https://openalex.org/C2777340749","wikidata":"https://www.wikidata.org/wiki/Q6684955","display_name":"Lottery","level":2,"score":0.44449999928474426},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.44290000200271606},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.4352000057697296},{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.42980000376701355},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.41519999504089355},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.38600000739097595},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35359999537467957},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.3433000147342682},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.32820001244544983},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3276999890804291},{"id":"https://openalex.org/C37404715","wikidata":"https://www.wikidata.org/wiki/Q380679","display_name":"Dynamic programming","level":2,"score":0.3050000071525574},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.2939999997615814},{"id":"https://openalex.org/C536738050","wikidata":"https://www.wikidata.org/wiki/Q3249071","display_name":"Social Welfare","level":2,"score":0.2930999994277954},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.28769999742507935},{"id":"https://openalex.org/C46814582","wikidata":"https://www.wikidata.org/wiki/Q23389","display_name":"Nash equilibrium","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.2639999985694885},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C2779629923","wikidata":"https://www.wikidata.org/wiki/Q1754086","display_name":"Social welfare function","level":3,"score":0.25929999351501465},{"id":"https://openalex.org/C79416737","wikidata":"https://www.wikidata.org/wiki/Q2305519","display_name":"Social learning","level":2,"score":0.25609999895095825},{"id":"https://openalex.org/C2778445095","wikidata":"https://www.wikidata.org/wiki/Q18354077","display_name":"Sample complexity","level":2,"score":0.25220000743865967}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.05450","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05450","pdf_url":"https://arxiv.org/pdf/2610.05450","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.2610.05450","doi":"https://doi.org/10.48550/arxiv.2610.05450","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.05450","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:2610.05450","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05450","pdf_url":"https://arxiv.org/pdf/2610.05450","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":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7220719053.pdf","grobid_xml":"https://content.openalex.org/works/W7220719053.grobid-xml"},"referenced_works_count":16,"referenced_works":["https://openalex.org/W1530616289","https://openalex.org/W1966938374","https://openalex.org/W2013784666","https://openalex.org/W2069317438","https://openalex.org/W2083459869","https://openalex.org/W2120199131","https://openalex.org/W2295870642","https://openalex.org/W2974490848","https://openalex.org/W3134055822","https://openalex.org/W4283168526","https://openalex.org/W4293404332","https://openalex.org/W4392677981","https://openalex.org/W4396601201","https://openalex.org/W4417115501","https://openalex.org/W7148604237","https://openalex.org/W7213028900"],"related_works":["https://openalex.org/W4416613238","https://openalex.org/W7213028900","https://openalex.org/W4400025421","https://openalex.org/W4414959919","https://openalex.org/W7135098070","https://openalex.org/W4391835636","https://openalex.org/W4414845277","https://openalex.org/W7102393664","https://openalex.org/W7151575003","https://openalex.org/W7197026380"],"abstract_inverted_index":{"Preference-based":[0],"alignment":[1],"methods":[2],"such":[3],"as":[4],"reinforcement":[5],"learning":[6,13],"from":[7,14,112],"human":[8,15,178],"feedback":[9,16],"(RLHF)":[10],"and":[11,45,67,92,149,181,192,200],"Nash":[12],"(NLHF)":[17],"aggregate":[18],"pairwise":[19],"preferences":[20,163],"to":[21,97,198],"learn":[22],"an":[23,102],"LLM":[24],"policy,":[25],"but":[26],"a":[27,108,134,171],"natural":[28],"goal":[29],"is":[30,75],"maximizing":[31],"social":[32],"welfare":[33,58],"(average":[34],"cardinal":[35],"utility),":[36],"which":[37],"comparisons":[38],"alone":[39],"need":[40],"not":[41],"identify.":[42],"G\u00f6lz,":[43],"Haghtalab,":[44],"Yang":[46],"(GHY)":[47],"measure":[48],"the":[49,53,57,60,69,81,145,152],"gap":[50],"by":[51],"distortion:":[52],"worst-case":[54],"ratio":[55],"between":[56],"of":[59,68,147,166],"best":[61],"fixed":[62],"lottery":[63],"(distribution":[64],"over":[65],"responses)":[66],"learned":[70],"lottery.":[71,83],"They":[72],"show":[73],"NLHF":[74,199],"optimal":[76,125],"when":[77,168],"every":[78,131,189],"user":[79],"receives":[80],"same":[82],"Account-based":[84],"LLMs,":[85],"however,":[86],"have":[87],"information":[88],"about":[89],"their":[90],"users":[91],"can":[93],"serve":[94],"different":[95,98],"lotteries":[96],"people.":[99],"We":[100],"give":[101],"efficient":[103],"algorithm,":[104],"GLHF,":[105],"that":[106],"learns":[107],"single":[109],"group-conditioned":[110],"policy":[111],"one":[113,167],"comparison":[114],"per":[115],"user.":[116],"Under":[117],"individual":[118],"Bradley--Terry":[119],"comparisons,":[120],"GLHF":[121,185],"asymptotically":[122],"matches":[123],"GHY's":[124],"population":[126],"distortion":[127,165,187,196],"bound":[128],"simultaneously":[129],"on":[130],"group":[132,154,191],"in":[133,144,188],"prespecified,":[135],"possibly":[136],"overlapping":[137],"collection,":[138],"with":[139,151,161],"sample":[140],"complexity":[141],"growing":[142],"logarithmically":[143],"number":[146],"groups":[148,160],"inversely":[150],"smallest":[153],"mass.":[155],"A":[156],"sharper":[157],"guarantee":[158],"for":[159],"similar":[162],"approaches":[164],"members":[169],"share":[170],"feasible":[172],"favorite":[173],"response.":[174],"In":[175],"experiments":[176],"using":[177],"coffee":[179],"ratings":[180],"synthetic":[182],"LLM-generated":[183],"ratings,":[184],"lowers":[186],"evaluated":[190],"substantially":[193],"reduces":[194],"worst-group":[195],"relative":[197],"other":[201],"group-agnostic":[202],"baselines.":[203]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-07T00:00:00"}