{"id":"https://openalex.org/W7166841412","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1184","title":"Sem-DPO: Mitigating Semantic Inconsistency in Preference Optimization for Prompt Engineering","display_name":"Sem-DPO: Mitigating Semantic Inconsistency in Preference Optimization for Prompt Engineering","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166841412","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1184"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.1184","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1184","pdf_url":"https://aclanthology.org/2026.findings-acl.1184.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.1184.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139807585","display_name":"Anas Mohamed","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anas Mohamed","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063807294","display_name":"Azal Ahmad Khan","orcid":"https://orcid.org/0009-0000-9435-5328"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Azal Ahmad Khan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139753670","display_name":"Xinran Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinran Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139716416","display_name":"Ahmad Faraz Khan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ahmad Faraz Khan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139832870","display_name":"Shuwen Ge","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuwen Ge","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139816056","display_name":"Saman Bahzad Khan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saman Bahzad Khan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024921948","display_name":"Ayaan Ahmad","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ayaan Ahmad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139816374","display_name":"Ali Anwar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ali Anwar","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.85524964,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"23656","last_page":"23674"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.746999979019165,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.746999979019165,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.026900000870227814,"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/T10848","display_name":"Advanced Multi-Objective Optimization Algorithms","score":0.018400000408291817,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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":0.4794999957084656},{"id":"https://openalex.org/keywords/dynamic-inconsistency","display_name":"Dynamic inconsistency","score":0.3151000142097473},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.2689000070095062},{"id":"https://openalex.org/keywords/probabilistic-logic","display_name":"Probabilistic logic","score":0.2551000118255615},{"id":"https://openalex.org/keywords/term","display_name":"Term (time)","score":0.25060001015663147}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5791000127792358},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.4794999957084656},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4296000003814697},{"id":"https://openalex.org/C127729010","wikidata":"https://www.wikidata.org/wiki/Q60165","display_name":"Dynamic inconsistency","level":2,"score":0.3151000142097473},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.29319998621940613},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.26499998569488525},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.25060001015663147},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.24690000712871552}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.1184","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1184","pdf_url":"https://aclanthology.org/2026.findings-acl.1184.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.1184","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1184","pdf_url":"https://aclanthology.org/2026.findings-acl.1184.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G4293822378","display_name":"PDaSP Track 3: Testbed for Enhancing Privacy and Robustness of Federated Learning Systems","funder_award_id":"2452817","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320332195","display_name":"Samsung","ror":"https://ror.org/04w3jy968"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166841412.pdf","grobid_xml":"https://content.openalex.org/works/W7166841412.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Generative":[0],"AI":[1],"can":[2,50],"now":[3],"synthesize":[4],"strikingly":[5],"realistic":[6],"images":[7],"from":[8,54,90],"text,":[9],"yet":[10,69],"output":[11],"quality":[12],"remains":[13],"highly":[14],"sensitive":[15],"to":[16,29],"how":[17,84],"prompts":[18,44,120],"are":[19,100],"phrased.Direct":[20],"Preference":[21],"Optimization":[22],"(DPO)":[23],"offers":[24],"a":[25,61,80,122],"lightweight,":[26],"off-policy":[27],"alternative":[28],"RL":[30],"for":[31,111,181,188],"automatic":[32],"prompt":[33,88,113,157],"engineering,":[34],"but":[35],"its":[36,71],"token-level":[37],"regularization":[38],"leaves":[39],"semantic":[40,67,109,174],"inconsistency":[41],"unchecked":[42],"as":[43,160,162],"that":[45,65,99,116,168],"win":[46],"higher":[47,142,147],"preference":[48,191],"scores":[49,149],"still":[51],"drift":[52,110],"away":[53],"the":[55,76,86,91,94,104,127,178,186],"user's":[56],"intended":[57],"meaning.We":[58],"introduce":[59],"Sem-DPO,":[60],"variant":[62],"of":[63,96,126],"DPO":[64,77,164],"preserves":[66],"consistency":[68],"retains":[70],"simplicity":[72],"and":[73,135,145,184],"efficiency.Sem-DPO":[74],"adjusts":[75],"loss":[78],"using":[79],"weight":[81],"based":[82],"on":[83,108],"different":[85],"winning":[87],"is":[89],"original,":[92],"reducing":[93],"impact":[95],"training":[97],"examples":[98],"semantically":[101],"misaligned.We":[102],"provide":[103],"first":[105],"analytical":[106],"bound":[107],"preference-tuned":[112],"generators,":[114],"showing":[115],"Sem-DPO":[117,139],"keeps":[118],"learned":[119],"within":[121],"provably":[123],"bounded":[124],"neighborhood":[125],"original":[128],"text.On":[129],"three":[130,136],"standard":[131,180],"text-to-image":[132],"prompt-optimization":[133,182],"benchmarks":[134],"language":[137,194],"models,":[138],"achieves":[140],"8-12%":[141],"CLIP":[143],"similarity":[144],"5-9%":[146],"human-preference":[148],"(HPSv2.1,PickScore)":[150],"than":[151],"DPO,":[152],"while":[153],"also":[154],"outperforming":[155],"state-ofthe-art":[156],"optimization":[158,192],"baselines":[159,171],"well":[161],"several":[163],"variants.These":[165],"findings":[166],"suggest":[167],"strong":[169],"flat":[170],"augmented":[172],"with":[173],"weighting":[175],"should":[176],"become":[177],"new":[179],"studies":[183],"lay":[185],"groundwork":[187],"broader,":[189],"semantics-aware":[190],"in":[193],"models.":[195]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}