{"id":"https://openalex.org/W7204896979","doi":"https://doi.org/10.48550/arxiv.2608.31079","title":"Sycophantic Agreement Transfers with Neutral Data via Contrastive Preference Optimization","display_name":"Sycophantic Agreement Transfers with Neutral Data via Contrastive Preference Optimization","publication_year":2026,"publication_date":"2026-08-31","ids":{"openalex":"https://openalex.org/W7204896979","doi":"https://doi.org/10.48550/arxiv.2608.31079"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.31079","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.31079","pdf_url":"https://arxiv.org/pdf/2608.31079","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.31079","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137653587","display_name":"Camila Blank","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Blank, Camila","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033138733","display_name":"Zhuofan Ying","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ying, Zhuofan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042601761","display_name":"Christopher Potts","orcid":"https://orcid.org/0000-0002-7978-6055","observed_orcids":["https://orcid.org/0000-0002-7978-6055","https://orcid.org/0009-0003-5415-040X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Potts, Christopher","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092574633","display_name":"Peter Hase","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hase, Peter","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5148839362","display_name":"Jing Huang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Jing","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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.3490999937057495,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.3490999937057495,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"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":0.11949999630451202,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.11349999904632568,"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-data","display_name":"preference data","score":0.5619999766349792},{"id":"https://openalex.org/keywords/model-alignment","display_name":"model alignment","score":0.6539999842643738},{"id":"https://openalex.org/keywords/data-attribution","display_name":"data attribution","score":0.49000000953674316},{"id":"https://openalex.org/keywords/factual-accuracy","display_name":"factual accuracy","score":0.5370000004768372}],"concepts":[{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.7820000052452087},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5491999983787537},{"id":"https://openalex.org/C2776818064","wikidata":"https://www.wikidata.org/wiki/Q829903","display_name":"Agreement","level":2,"score":0.5472000241279602},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.49950000643730164},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.4722000062465668},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.4519999921321869},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.44609999656677246},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43459999561309814},{"id":"https://openalex.org/C2779110102","wikidata":"https://www.wikidata.org/wiki/Q1323737","display_name":"Revealed preference","level":2,"score":0.3937999904155731},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.38960000872612},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3587999939918518},{"id":"https://openalex.org/C2777868144","wikidata":"https://www.wikidata.org/wiki/Q7239817","display_name":"Preference elicitation","level":3,"score":0.34599998593330383},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.336899995803833},{"id":"https://openalex.org/C117220453","wikidata":"https://www.wikidata.org/wiki/Q5172842","display_name":"Correlation","level":2,"score":0.33550000190734863},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33169999718666077},{"id":"https://openalex.org/C55037315","wikidata":"https://www.wikidata.org/wiki/Q5421151","display_name":"Experimental data","level":2,"score":0.29809999465942383},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.29440000653266907},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.28999999165534973},{"id":"https://openalex.org/C2776175482","wikidata":"https://www.wikidata.org/wiki/Q1195816","display_name":"Transfer (computing)","level":2,"score":0.2874000072479248},{"id":"https://openalex.org/C143299363","wikidata":"https://www.wikidata.org/wiki/Q900584","display_name":"Attribution","level":2,"score":0.27900001406669617},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2786000072956085},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C64357122","wikidata":"https://www.wikidata.org/wiki/Q1149766","display_name":"Causality (physics)","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.25850000977516174},{"id":"https://openalex.org/C59656382","wikidata":"https://www.wikidata.org/wiki/Q191536","display_name":"Conjunction (astronomy)","level":2,"score":0.2565999925136566}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.31079","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.31079","pdf_url":"https://arxiv.org/pdf/2608.31079","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.2608.31079","doi":"https://doi.org/10.48550/arxiv.2608.31079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.31079","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:2608.31079","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.31079","pdf_url":"https://arxiv.org/pdf/2608.31079","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7214999567","https://openalex.org/W7127436076","https://openalex.org/W7166832698","https://openalex.org/W7203687722","https://openalex.org/W7130941786","https://openalex.org/W4417114736","https://openalex.org/W4416036648","https://openalex.org/W4403564074","https://openalex.org/W7151890762","https://openalex.org/W4415958555"],"abstract_inverted_index":{"Sycophantic":[0],"agreement":[1,23,49,94,102],"refers":[2],"to":[3,114,133,187,207,221],"a":[4,25,83,158,192],"behavior":[5],"in":[6,157,217],"which":[7],"language":[8],"models":[9,77,205],"excessively":[10],"affirm":[11],"the":[12,16,64,87,90,97,139,145,151,196,203],"user,":[13],"often":[14],"at":[15],"cost":[17],"of":[18,28,35,56,75,89,161,173,195],"factual":[19],"accuracy.":[20],"Although":[21],"sycophantic":[22,48,93,101,174,228],"is":[24,32,82,111,148],"well-known":[26],"failure":[27],"model":[29,40,92,100],"alignment,":[30],"there":[31,81,168],"limited":[33,113],"understanding":[34],"how":[36],"it":[37],"emerges":[38],"from":[39],"training.":[41],"In":[42],"this":[43,108,128],"work,":[44],"we":[45,69,137],"demonstrate":[46,106],"that":[47,107,144,202],"can":[50,130,211],"emerge":[51],"as":[52],"an":[53],"unintended":[54,109],"consequence":[55],"widely":[57],"used":[58,206],"contrastive":[59],"preference":[60,122,140,209],"optimization":[61,123],"objectives.":[62,124],"Using":[63],"OLMo":[65],"3":[66],"post-training":[67],"pipeline,":[68],"show":[70],"that,":[71],"for":[72],"various":[73],"pairs":[74],"teacher":[76,91,204],"across":[78,119,150],"three":[79],"families,":[80],"strong":[84],"correlation":[85],"between":[86],"log-ratio":[88],"rates":[95],"and":[96,142,176,223],"resulting":[98],"student":[99],"rate.":[103],"We":[104],"further":[105],"transfer":[110],"not":[112],"DPO":[115],"but":[116],"also":[117],"occurs":[118],"6":[120],"other":[121],"To":[125],"understand":[126],"whether":[127],"effect":[129],"be":[131],"attributed":[132],"particular":[134],"training":[135,215],"examples,":[136],"analyze":[138],"data":[141,181,210],"find":[143],"sycophancy":[146,189],"signal":[147],"diffused":[149],"entire":[152],"dataset":[153],"rather":[154],"than":[155],"concentrated":[156],"sparse":[159],"set":[160],"examples:":[162],"each":[163],"example":[164],"appears":[165],"neutral,":[166],"i.e.,":[167],"are":[169],"no":[170],"explicit":[171],"instances":[172],"agreement,":[175],"filtering":[177],"based":[178],"on":[179],"probe-based":[180],"attribution":[182],"or":[183],"logit-linear":[184],"selection":[185],"fails":[186],"mitigate":[188],"without":[190],"removing":[191],"large":[193],"portion":[194],"dataset.":[197],"Overall,":[198],"our":[199],"findings":[200],"suggest":[201],"generate":[208],"interact":[212],"with":[213],"alignment":[214],"objectives":[216],"unexpected":[218],"ways,":[219],"generalizing":[220],"undesirable":[222],"potentially":[224],"harmful":[225],"behaviors":[226],"like":[227],"agreement.":[229]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-02T00:00:00"}