{"id":"https://openalex.org/W7138974071","doi":"https://doi.org/10.1609/aaai.v40i44.41143","title":"When Human Preferences Flip: An Instance-Dependent Robust Loss for RLHF","display_name":"When Human Preferences Flip: An Instance-Dependent Robust Loss for RLHF","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138974071","doi":"https://doi.org/10.1609/aaai.v40i44.41143"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i44.41143","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i44.41143","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/41143/45104","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://ojs.aaai.org/index.php/AAAI/article/download/41143/45104","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101309972","display_name":"Yifan Xu","orcid":"https://orcid.org/0009-0003-7798-647X"},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Yifan Xu","raw_affiliation_strings":["Hong Kong Baptist University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129970956","display_name":"Xichen Ye","orcid":null},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xichen Ye","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Yifan Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Yifan Chen","raw_affiliation_strings":["Hong Kong Baptist University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5085826827","display_name":"Qiang Zhang","orcid":"https://orcid.org/0000-0001-6662-1668"},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4391012619","display_name":"Shanghai Artificial Intelligence Laboratory","ror":"https://ror.org/03wkvpx79","country_code":null,"type":"facility","lineage":["https://openalex.org/I4391012619"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qiaosheng Zhang","raw_affiliation_strings":["Shanghai Artificial Intelligence Laboratory"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Artificial Intelligence Laboratory","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4391012619"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.75071225,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"44","first_page":"38057","last_page":"38065"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.49160000681877136,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.49160000681877136,"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/T10028","display_name":"Topic Modeling","score":0.15479999780654907,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.03280000016093254,"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/preference","display_name":"Preference","score":0.7229999899864197},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.6186000108718872},{"id":"https://openalex.org/keywords/preference-learning","display_name":"Preference learning","score":0.5429999828338623},{"id":"https://openalex.org/keywords/preference-elicitation","display_name":"Preference elicitation","score":0.4262999892234802},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.3386000096797943},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.32919999957084656},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.3278999924659729}],"concepts":[{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.7229999899864197},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6456000208854675},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6186000108718872},{"id":"https://openalex.org/C181204326","wikidata":"https://www.wikidata.org/wiki/Q7239820","display_name":"Preference learning","level":3,"score":0.5429999828338623},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5407999753952026},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4528000056743622},{"id":"https://openalex.org/C2777868144","wikidata":"https://www.wikidata.org/wiki/Q7239817","display_name":"Preference elicitation","level":3,"score":0.4262999892234802},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3386000096797943},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.32919999957084656},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C32172795","wikidata":"https://www.wikidata.org/wiki/Q4692266","display_name":"Aggregation problem","level":2,"score":0.30820000171661377},{"id":"https://openalex.org/C193254401","wikidata":"https://www.wikidata.org/wiki/Q2160088","display_name":"Robust optimization","level":2,"score":0.30820000171661377},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.30250000953674316},{"id":"https://openalex.org/C159694833","wikidata":"https://www.wikidata.org/wiki/Q2321565","display_name":"Iterative method","level":2,"score":0.27959999442100525},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C2987595161","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Optimization algorithm","level":2,"score":0.26739999651908875},{"id":"https://openalex.org/C2781043087","wikidata":"https://www.wikidata.org/wiki/Q939761","display_name":"Preference theory","level":3,"score":0.26510000228881836},{"id":"https://openalex.org/C117619785","wikidata":"https://www.wikidata.org/wiki/Q6094414","display_name":"Iterative learning control","level":3,"score":0.2635999917984009}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i44.41143","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i44.41143","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/41143/45104","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i44.41143","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i44.41143","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/41143/45104","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.8059132695198059}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7138974071.pdf","grobid_xml":"https://content.openalex.org/works/W7138974071.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Quality":[0],"of":[1,99,159],"datasets":[2],"plays":[3],"an":[4,92],"important":[5],"role":[6],"in":[7,25,87,115],"large":[8],"language":[9],"model":[10,73,118,153],"(LLM)":[11],"alignment.":[12],"In":[13,122,144],"collecting":[14],"human":[15,63,71],"feedback,":[16],"however,":[17],"preference":[18,56,76,110,119,151],"flipping":[19,57,77,94,120,152],"is":[20],"ubiquitous":[21],"and":[22,74,117,138],"causes":[23],"corruption":[24],"data":[26],"annotation;":[27],"the":[28,31,69,75,88,97,100,135,149],"issue":[29],"necessitates":[30],"alignment":[32],"algorithms":[33],"with":[34,62,134],"improved":[35],"robustness":[36],"against":[37],"potential":[38],"flipped":[39],"pairs.":[40],"To":[41],"this":[42,44],"end,":[43],"paper":[45],"introduces":[46],"a":[47,59,126],"Flipping-Aware":[48],"Direct":[49,139],"Preference":[50,140],"Optimization":[51,141],"(FA-DPO)":[52],"algorithm":[53,132],"tailored":[54],"to":[55,109],"from":[58],"reinforcement":[60],"learning":[61],"feedback":[64],"(RLHF)":[65],"perspective.":[66],"We":[67],"dissect":[68],"inherent":[70],"intention":[72],"mechanism":[78],"introduced":[79],"by":[80,105],"external":[81],"factors":[82],"as":[83,163,165],"two":[84],"distinct":[85],"stages;":[86],"latter,":[89],"we":[90,112,124,147],"introduce":[91],"instance-dependent":[93,150],"probability":[95],"on":[96],"basis":[98],"Bradley-Terry":[101],"(BT)":[102],"model.":[103],"Further,":[104],"leveraging":[106],"features":[107],"relevant":[108],"annotation,":[111],"capture":[113],"uncertainty":[114],"judgments":[116],"patterns.":[121],"practice,":[123],"design":[125],"simple":[127],"yet":[128],"efficient":[129],"iterative":[130],"optimization":[131],"compatible":[133],"original":[136],"RLHF":[137],"(DPO)":[142],"algorithms.":[143],"our":[145,160],"experiments,":[146],"investigate":[148],"under":[154],"multiple":[155],"circumstances":[156],"for":[157],"evaluation":[158],"proposed":[161],"method,":[162],"well":[164],"other":[166],"baseline":[167],"methods.":[168]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-03-20T00:00:00"}