{"id":"https://openalex.org/W7166183413","doi":"https://doi.org/10.48550/arxiv.2606.27373","title":"Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models","display_name":"Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models","publication_year":2026,"publication_date":"2026-06-25","ids":{"openalex":"https://openalex.org/W7166183413","doi":"https://doi.org/10.48550/arxiv.2606.27373"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.27373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.27373","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.27373","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139402485","display_name":"Shravan Venkatraman","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Venkatraman, Shravan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115860474","display_name":"Ritesh Thawkar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thawkar, Ritesh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134365995","display_name":"Omkar Thawakar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thawakar, Omkar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139394009","display_name":"Rao Muhammad Anwer","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anwer, Rao Muhammad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009362997","display_name":"Hisham Cholakkal","orcid":"https://orcid.org/0000-0002-8230-9065"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cholakkal, Hisham","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139408770","display_name":"Salman Khan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khan, Salman","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139439585","display_name":"Fahad Khan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khan, Fahad","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9872000217437744,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9872000217437744,"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/T10028","display_name":"Topic Modeling","score":0.0038999998942017555,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0020000000949949026,"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/prior-probability","display_name":"Prior probability","score":0.6259999871253967},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.5444999933242798},{"id":"https://openalex.org/keywords/closed-captioning","display_name":"Closed captioning","score":0.5437999963760376},{"id":"https://openalex.org/keywords/image","display_name":"Image (mathematics)","score":0.45809999108314514},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.43959999084472656},{"id":"https://openalex.org/keywords/object","display_name":"Object (grammar)","score":0.436599999666214},{"id":"https://openalex.org/keywords/unsupervised-learning","display_name":"Unsupervised learning","score":0.40639999508857727},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4016999900341034},{"id":"https://openalex.org/keywords/visual-reasoning","display_name":"Visual reasoning","score":0.34630000591278076},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.33079999685287476}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6800000071525574},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.6259999871253967},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6248999834060669},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.5444999933242798},{"id":"https://openalex.org/C157657479","wikidata":"https://www.wikidata.org/wiki/Q2367247","display_name":"Closed captioning","level":3,"score":0.5437999963760376},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.45809999108314514},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.43959999084472656},{"id":"https://openalex.org/C2781238097","wikidata":"https://www.wikidata.org/wiki/Q175026","display_name":"Object (grammar)","level":2,"score":0.436599999666214},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.42080000042915344},{"id":"https://openalex.org/C8038995","wikidata":"https://www.wikidata.org/wiki/Q1152135","display_name":"Unsupervised learning","level":2,"score":0.40639999508857727},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4016999900341034},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.34630000591278076},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.33079999685287476},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.3260999917984009},{"id":"https://openalex.org/C64876066","wikidata":"https://www.wikidata.org/wiki/Q5141226","display_name":"Cognitive neuroscience of visual object recognition","level":3,"score":0.3199999928474426},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.31709998846054077},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.31679999828338623},{"id":"https://openalex.org/C2776036281","wikidata":"https://www.wikidata.org/wiki/Q48769818","display_name":"Constraint (computer-aided design)","level":2,"score":0.31529998779296875},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.31040000915527344},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3095000088214874},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3077999949455261},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.3066999912261963},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.303600013256073},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.2912999987602234},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.28529998660087585},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.28529998660087585},{"id":"https://openalex.org/C2780103172","wikidata":"https://www.wikidata.org/wiki/Q1309721","display_name":"Visual Objects","level":3,"score":0.2799000144004822},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2786000072956085},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2655999958515167},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C2776151529","wikidata":"https://www.wikidata.org/wiki/Q3045304","display_name":"Object detection","level":3,"score":0.25279998779296875}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.27373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.27373","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.27373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.27373","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":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recently,":[0],"self-evolving":[1,27,86,113],"large":[2],"multimodal":[3],"models":[4,231],"(LMMs)":[5],"have":[6],"received":[7],"attention":[8,78],"for":[9],"improving":[10],"visual":[11,38,60,80,98,120],"reasoning":[12],"in":[13,25],"a":[14,54,83,110,128,140,166],"purely":[15,111],"unsupervised":[16,112],"setting.":[17],"However,":[18],"multi-role":[19],"self-play":[20],"and":[21,97,139,177,208,220,226,230],"self-consistency":[22],"reward":[23,131,143,173],"schemes":[24],"existing":[26],"LMMs":[28,87],"optimize":[29],"answer":[30],"agreement":[31],"without":[32,169],"ensuring":[33],"the":[34,63,71,118,150,154,189,197],"decoder":[35,64],"attends":[36],"to":[37,46,53,79,152],"content,":[39],"relying":[40],"instead":[41],"on":[42,66,89,180,185,206,211],"statistical":[43],"language":[44,67],"priors":[45,68],"produce":[47],"self":[48],"consistent":[49],"outputs.":[50],"This":[51],"leads":[52],"persistent":[55],"failure":[56],"mode":[57],"we":[58,104],"term":[59],"under-conditioning,":[61],"where":[62],"relies":[65],"rather":[69],"than":[70],"image":[72,95],"during":[73],"generation,":[74],"manifesting":[75],"as":[76,94,196],"insufficient":[77],"tokens.":[81],"As":[82],"result,":[84],"current":[85],"struggle":[88],"vision--language":[90],"understanding":[91],"tasks":[92],"such":[93],"captioning":[96],"question":[99],"answering.":[100],"To":[101],"address":[102],"this,":[103],"propose":[105],"VISE":[106,163,200],"(Visual":[107],"Invariance":[108],"Self-Evolution),":[109],"framework":[114],"that":[115,132,144],"directly":[116],"regularizes":[117],"model's":[119],"conditioning":[121],"policy":[122],"through":[123],"two":[124],"complementary":[125],"invariance-based":[126],"rewards:":[127],"geometric":[129],"invariance":[130,142],"enforces":[133],"spatial":[134],"consistency":[135],"under":[136],"known":[137],"transformations,":[138],"semantic":[141],"penalizes":[145],"evidence-agnostic":[146],"generation":[147],"by":[148,216],"requiring":[149],"model":[151,168,224],"recognize":[153],"absence":[155],"of":[156,191,203],"evidence":[157],"when":[158],"predicted":[159],"regions":[160],"are":[161,232],"perturbed.":[162],"operates":[164],"within":[165],"single":[167],"specialist":[170],"roles,":[171],"external":[172],"models,":[174],"or":[175],"annotations,":[176],"is":[178],"trained":[179],"raw":[181],"unlabeled":[182],"images.":[183],"Experiments":[184],"18":[186],"benchmarks":[187],"demonstrate":[188],"efficacy":[190],"our":[192],"approach.":[193],"Using":[194],"Qwen3-VL-2B":[195],"base":[198],"model,":[199],"achieves":[201],"gains":[202],"$+16.85$":[204],"CIDEr":[205,210],"COCO":[207],"$+19.66$":[209],"TextCaps,":[212],"reduces":[213],"object":[214],"hallucination":[215],"$5.0$":[217],"Chair-I":[218],"points,":[219],"generalizes":[221],"across":[222],"four":[223],"families":[225],"scales.":[227],"Our":[228],"code":[229],"available":[233],"at":[234],"https://mbzuai-oryx.github.io/VISE":[235]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-27T00:00:00"}