{"id":"https://openalex.org/W7138255619","doi":"https://doi.org/10.1609/aaai.v40i12.38000","title":"ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLM","display_name":"ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLM","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138255619","doi":"https://doi.org/10.1609/aaai.v40i12.38000"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i12.38000","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i12.38000","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/38000/41962","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/38000/41962","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100362598","display_name":"Yujun Wang","orcid":"https://orcid.org/0000-0002-0921-0122"},"institutions":[{"id":"https://openalex.org/I32021983","display_name":"Christian-Albrechts-Universit\u00e4t zu Kiel","ror":"https://ror.org/04v76ef78","country_code":"DE","type":"education","lineage":["https://openalex.org/I32021983"]}],"countries":["DE"],"is_corresponding":true,"raw_author_name":"Yujun Wang","raw_affiliation_strings":["Christian-Albrechts-Universit\u00e4t zu Kiel (CAU)"],"affiliations":[{"raw_affiliation_string":"Christian-Albrechts-Universit\u00e4t zu Kiel (CAU)","institution_ids":["https://openalex.org/I32021983"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129731231","display_name":"Aniri","orcid":null},"institutions":[{"id":"https://openalex.org/I8204097","display_name":"Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen","ror":"https://ror.org/05591te55","country_code":"DE","type":"education","lineage":["https://openalex.org/I8204097"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Aniri","raw_affiliation_strings":["Technical University of Munich\nLudwig Maximilian University of Munich"],"affiliations":[{"raw_affiliation_string":"Technical University of Munich\nLudwig Maximilian University of Munich","institution_ids":["https://openalex.org/I8204097"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129712780","display_name":"Jinhe Bi","orcid":null},"institutions":[{"id":"https://openalex.org/I8204097","display_name":"Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen","ror":"https://ror.org/05591te55","country_code":"DE","type":"education","lineage":["https://openalex.org/I8204097"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Jinhe Bi","raw_affiliation_strings":["Ludwig Maximilian University of Munich\nMunich Center for Machine Learning"],"affiliations":[{"raw_affiliation_string":"Ludwig Maximilian University of Munich\nMunich Center for Machine Learning","institution_ids":["https://openalex.org/I8204097"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065911487","display_name":"S\u00f6ren Pirk","orcid":null},"institutions":[{"id":"https://openalex.org/I32021983","display_name":"Christian-Albrechts-Universit\u00e4t zu Kiel","ror":"https://ror.org/04v76ef78","country_code":"DE","type":"education","lineage":["https://openalex.org/I32021983"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Soren Pirk","raw_affiliation_strings":["Christian-Albrechts-Universit\u00e4t zu Kiel (CAU)"],"affiliations":[{"raw_affiliation_string":"Christian-Albrechts-Universit\u00e4t zu Kiel (CAU)","institution_ids":["https://openalex.org/I32021983"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5124229701","display_name":"Yunpu Ma","orcid":null},"institutions":[{"id":"https://openalex.org/I8204097","display_name":"Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen","ror":"https://ror.org/05591te55","country_code":"DE","type":"education","lineage":["https://openalex.org/I8204097"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Yunpu Ma","raw_affiliation_strings":["Ludwig Maximilian University of Munich\nMunich Center for Machine Learning"],"affiliations":[{"raw_affiliation_string":"Ludwig Maximilian University of Munich\nMunich Center for Machine Learning","institution_ids":["https://openalex.org/I8204097"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5100362598"],"corresponding_institution_ids":["https://openalex.org/I32021983"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.46895075,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"12","first_page":"10306","last_page":"10314"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.20340000092983246,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.20340000092983246,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.1453000009059906,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.08730000257492065,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/decoding-methods","display_name":"Decoding methods","score":0.8256999850273132},{"id":"https://openalex.org/keywords/hallucinating","display_name":"Hallucinating","score":0.7639999985694885},{"id":"https://openalex.org/keywords/spurious-relationship","display_name":"Spurious relationship","score":0.7254999876022339},{"id":"https://openalex.org/keywords/mechanism","display_name":"Mechanism (biology)","score":0.40119999647140503},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.36090001463890076},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.31850001215934753}],"concepts":[{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.8256999850273132},{"id":"https://openalex.org/C2911011789","wikidata":"https://www.wikidata.org/wiki/Q130741","display_name":"Hallucinating","level":2,"score":0.7639999985694885},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.7254999876022339},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7069000005722046},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5656999945640564},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.4343000054359436},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.40119999647140503},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.36410000920295715},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.36090001463890076},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3384000062942505},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.31850001215934753},{"id":"https://openalex.org/C198082294","wikidata":"https://www.wikidata.org/wiki/Q3399648","display_name":"Position (finance)","level":2,"score":0.3077999949455261},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.28690001368522644},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.27129998803138733},{"id":"https://openalex.org/C2781449363","wikidata":"https://www.wikidata.org/wiki/Q156751","display_name":"Anaphora (linguistics)","level":3,"score":0.26840001344680786},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.2619999945163727}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i12.38000","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i12.38000","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/38000/41962","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.v40i12.38000","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i12.38000","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/38000/41962","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/4","score":0.5481345057487488,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7138255619.pdf","grobid_xml":"https://content.openalex.org/works/W7138255619.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1],"language":[2],"models":[3],"(MLLMs)":[4],"frequently":[5],"hallucinate":[6],"by":[7,115],"over-committing":[8],"to":[9,117,145],"spurious":[10],"visual":[11,86],"cues.":[12],"Prior":[13],"remedies\u2013Visual":[14],"and":[15,74,94,103,113,131,142],"Instruction":[16],"Contrastive":[17,49],"Decoding":[18,50],"(VCD,":[19],"ICD)\u2013mitigate":[20],"this":[21,44],"issue,":[22],"yet":[23],"the":[24,55],"mechanism":[25],"remains":[26],"opaque.":[27],"We":[28],"first":[29],"empirically":[30],"show":[31],"that":[32],"their":[33],"improvements":[34],"systematically":[35],"coincide":[36],"with":[37],"redistributions":[38],"of":[39],"cross-modal":[40],"attention.":[41],"Building":[42],"on":[43,110,122],"insight,":[45],"we":[46],"propose":[47],"Attention-Steerable":[48],"(ASCD),":[51],"which":[52,65,81],"directly":[53],"steers":[54],"attention":[56,136],"scores":[57],"during":[58],"decoding.":[59],"ASCD":[60,107],"combines":[61],"(i)":[62],"positive":[63],"steering,":[64,80],"amplifies":[66],"automatically":[67],"mined":[68],"text-centric":[69],"heads\u2013stable":[70],"within":[71],"a":[72,139],"model":[73],"robust":[75],"across":[76],"domains\u2013with":[77],"(ii)":[78],"negative":[79],"dampens":[82],"on-the-fly":[83],"identified":[84],"critical":[85],"tokens.":[87],"The":[88],"method":[89],"incurs":[90],"negligible":[91],"runtime/memory":[92],"overhead":[93],"requires":[95],"no":[96],"additional":[97],"training.":[98],"Across":[99],"five":[100],"MLLM":[101],"backbones":[102],"three":[104],"decoding":[105],"schemes,":[106],"reduces":[108],"hallucination":[109],"POPE,":[111],"CHAIR,":[112],"MMHal-Bench":[114],"up":[116],"38.2%":[118],"while":[119],"improving":[120],"accuracy":[121],"standard":[123],"VQA":[124],"benchmarks,":[125],"including":[126],"MMMU,":[127],"MM-VET,":[128],"ScienceQA,":[129],"TextVQA,":[130],"GQA.":[132],"These":[133],"results":[134],"position":[135],"steering":[137],"as":[138],"simple,":[140],"model-agnostic,":[141],"principled":[143],"route":[144],"safer,":[146],"more":[147],"faithful":[148],"multimodal":[149],"generation.":[150]},"counts_by_year":[],"updated_date":"2026-03-20T20:47:17.329874","created_date":"2026-03-18T00:00:00"}