{"id":"https://openalex.org/W7133246709","doi":"https://doi.org/10.48550/arxiv.2602.24041","title":"Look Carefully: Adaptive Visual Reinforcements in Multimodal Large Language Models for Hallucination Mitigation","display_name":"Look Carefully: Adaptive Visual Reinforcements in Multimodal Large Language Models for Hallucination Mitigation","publication_year":2026,"publication_date":"2026-02-27","ids":{"openalex":"https://openalex.org/W7133246709","doi":"https://doi.org/10.48550/arxiv.2602.24041"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2602.24041","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.24041","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2602.24041","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127808960","display_name":"Xingyu Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhu, Xingyu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124909181","display_name":"Kesen Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Kesen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127868822","display_name":"Liang Yi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yi, Liang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127814106","display_name":"Shuo Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Shuo","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127781427","display_name":"Zhicai Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhicai","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009738035","display_name":"Beier Zhu","orcid":"https://orcid.org/0000-0002-7900-6979"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Beier","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127847717","display_name":"Hanwang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Hanwang","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5127808960"],"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.15440000593662262,"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":0.15440000593662262,"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/T13397","display_name":"Hallucinations in medical conditions","score":0.15080000460147858,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.04859999939799309,"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/security-token","display_name":"Security token","score":0.59579998254776},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.593500018119812},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4693000018596649},{"id":"https://openalex.org/keywords/visualization","display_name":"Visualization","score":0.4490000009536743},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.4277999997138977},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.40070000290870667},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3562999963760376}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.718999981880188},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.59579998254776},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.593500018119812},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49810001254081726},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4693000018596649},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.4490000009536743},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.4277999997138977},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.40070000290870667},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3562999963760376},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.34929999709129333},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3215000033378601},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.3154999911785126},{"id":"https://openalex.org/C2986089797","wikidata":"https://www.wikidata.org/wiki/Q6501338","display_name":"Visual attention","level":3,"score":0.2904999852180481},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2831000089645386},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2816999852657318},{"id":"https://openalex.org/C32022120","wikidata":"https://www.wikidata.org/wiki/Q797225","display_name":"Interference (communication)","level":3,"score":0.2531000077724457},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2602.24041","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.24041","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2602.24041","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.24041","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10","score":0.49149295687675476},{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.4892963469028473}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1,101],"language":[2],"models":[3],"(MLLMs)":[4],"have":[5],"achieved":[6],"remarkable":[7],"progress":[8],"in":[9],"vision-language":[10],"reasoning,":[11],"yet":[12],"they":[13,57],"remain":[14],"vulnerable":[15],"to":[16,46,110,125],"hallucination,":[17],"where":[18],"generated":[19],"content":[20],"deviates":[21],"from":[22,66,73],"visual":[23,52,104,145],"evidence.":[24],"Existing":[25],"mitigation":[26],"strategies":[27],"either":[28],"require":[29],"costly":[30],"supervision":[31],"during":[32,54],"training":[33],"or":[34],"introduce":[35],"additional":[36],"latency":[37],"at":[38],"inference":[39],"time.":[40],"Recent":[41],"vision":[42],"enhancement":[43],"methods":[44],"attempt":[45],"address":[47],"this":[48,78],"issue":[49],"by":[50],"reinforcing":[51],"tokens":[53,61,105],"decoding,":[55],"but":[56],"typically":[58],"inject":[59],"all":[60],"indiscriminately,":[62],"which":[63],"causes":[64],"interference":[65],"background":[67],"regions":[68],"and":[69,122,147],"distracts":[70],"the":[71,100,117,128,140],"model":[72],"critical":[74],"cues.":[75],"To":[76],"overcome":[77],"challenge,":[79],"we":[80],"propose":[81],"Adaptive":[82],"Visual":[83],"Reinforcement":[84],"(AIR),":[85],"a":[86,107,136],"training-free":[87],"framework":[88],"for":[89,172],"MLLMs.":[90,175],"AIR":[91,138,158],"consists":[92],"of":[93,103],"two":[94],"components.":[95],"Prototype-based":[96],"token":[97],"reduction":[98],"condenses":[99],"pool":[102],"into":[106,132],"compact":[108],"subset":[109],"suppress":[111],"redundancy.":[112],"OT-guided":[113],"patch":[114,123],"reinforcement":[115],"quantifies":[116],"alignment":[118],"between":[119],"hidden":[120],"states":[121],"embeddings":[124],"selectively":[126],"integrate":[127],"most":[129],"consistent":[130],"patches":[131],"feed-forward":[133],"layers.":[134],"As":[135],"result,":[137],"enhances":[139],"model's":[141],"reliance":[142],"on":[143],"salient":[144],"information":[146],"effectively":[148],"mitigates":[149],"hallucination.":[150],"Extensive":[151],"experiments":[152],"across":[153],"representative":[154],"MLLMs":[155],"demonstrate":[156],"that":[157],"substantially":[159],"reduces":[160],"hallucination":[161],"while":[162],"preserving":[163],"general":[164],"capabilities,":[165],"establishing":[166],"it":[167],"as":[168],"an":[169],"effective":[170],"solution":[171],"building":[173],"reliable":[174]},"counts_by_year":[],"updated_date":"2026-03-03T06:18:10.843953","created_date":"2026-03-03T00:00:00"}