{"id":"https://openalex.org/W7155068464","doi":"https://doi.org/10.48550/arxiv.2604.17087","title":"EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling","display_name":"EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling","publication_year":2026,"publication_date":"2026-04-18","ids":{"openalex":"https://openalex.org/W7155068464","doi":"https://doi.org/10.48550/arxiv.2604.17087"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.17087","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17087","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.2604.17087","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134103080","display_name":"Jiafei Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Jiafei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052677272","display_name":"Fengwei Zhou","orcid":"https://orcid.org/0000-0002-6404-6441"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Fengwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134099808","display_name":"Jin Qu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qu, Jin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060160424","display_name":"Wenjin Jason Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Wenjin Jason","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134147667","display_name":"Tong Wu","orcid":"https://orcid.org/0000-0001-5848-7008"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Tong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110076836","display_name":"Gengjian Xue","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xue, Gengjian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134108822","display_name":"Zhikang Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Zhikang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134096213","display_name":"Daomin Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Daomin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134122076","display_name":"Yichao Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Yichao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5007397640","display_name":"Bailin Na","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Na, Bailin","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":0.9320999979972839,"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":0.9320999979972839,"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":0.015200000256299973,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.01489999983459711,"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/security-token","display_name":"Security token","score":0.7706999778747559},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5659999847412109},{"id":"https://openalex.org/keywords/cosine-similarity","display_name":"Cosine similarity","score":0.40799999237060547},{"id":"https://openalex.org/keywords/semantic-role-labeling","display_name":"Semantic role labeling","score":0.357699990272522},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3537999987602234},{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.35280001163482666},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.34220001101493835},{"id":"https://openalex.org/keywords/semantic-similarity","display_name":"Semantic similarity","score":0.328000009059906},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.32170000672340393}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8197000026702881},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7706999778747559},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5687999725341797},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5659999847412109},{"id":"https://openalex.org/C2780762811","wikidata":"https://www.wikidata.org/wiki/Q1784941","display_name":"Cosine similarity","level":3,"score":0.40799999237060547},{"id":"https://openalex.org/C67277372","wikidata":"https://www.wikidata.org/wiki/Q7449085","display_name":"Semantic role labeling","level":3,"score":0.357699990272522},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3537999987602234},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.35280001163482666},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.34220001101493835},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.328000009059906},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.32170000672340393},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3208000063896179},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3174999952316284},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3118000030517578},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.30480000376701355},{"id":"https://openalex.org/C35639132","wikidata":"https://www.wikidata.org/wiki/Q7452468","display_name":"Sequence labeling","level":3,"score":0.3034999966621399},{"id":"https://openalex.org/C2779668609","wikidata":"https://www.wikidata.org/wiki/Q623092","display_name":"Rapid serial visual presentation","level":3,"score":0.29789999127388},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.2757999897003174},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.26759999990463257},{"id":"https://openalex.org/C2221639","wikidata":"https://www.wikidata.org/wiki/Q2877","display_name":"Discrete cosine transform","level":3,"score":0.26409998536109924},{"id":"https://openalex.org/C159423971","wikidata":"https://www.wikidata.org/wiki/Q177251","display_name":"Associative property","level":2,"score":0.25929999351501465},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2578999996185303},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2508000135421753},{"id":"https://openalex.org/C4661277","wikidata":"https://www.wikidata.org/wiki/Q1054157","display_name":"Cog","level":2,"score":0.25029999017715454},{"id":"https://openalex.org/C2780103172","wikidata":"https://www.wikidata.org/wiki/Q1309721","display_name":"Visual Objects","level":3,"score":0.25}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.17087","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17087","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.2604.17087","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17087","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":[{"id":"https://metadata.un.org/sdg/4","score":0.7852014899253845,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"Multimodal":[1],"Large":[2],"Language":[3],"Models":[4],"(MLLMs)":[5],"have":[6],"demonstrated":[7],"strong":[8],"performance":[9],"on":[10,164,187],"vision-language":[11,155],"understanding":[12],"tasks,":[13],"yet":[14],"their":[15],"inference":[16],"efficiency":[17],"is":[18],"often":[19],"hampered":[20],"by":[21,71],"the":[22,64,88,105,120,128,174],"large":[23],"number":[24],"of":[25,173],"visual":[26,42,69,74],"tokens,":[27],"particularly":[28],"in":[29,82],"high-resolution":[30],"or":[31,166],"multi-image":[32],"scenarios.":[33],"To":[34,90],"address":[35],"this":[36,91],"issue,":[37],"we":[38,93],"propose":[39],"EvoComp,":[40],"a":[41,57,123,138],"token":[43,49,102,115,179],"compression":[44,180],"framework":[45],"that":[46,62,99,158],"significantly":[47],"reduces":[48],"count":[50],"while":[51,109],"preserving":[52],"task":[53],"accuracy.":[54],"EvoComp":[55,159],"introduces":[56],"lightweight":[58],"encoder-only":[59],"transformer-based":[60],"compressor":[61,121],"selects":[63],"most":[65],"informative":[66],"and":[67,75,134,137,148,181],"non-redundant":[68],"tokens":[70],"jointly":[72],"considering":[73],"textual":[76],"contexts.":[77],"A":[78],"core":[79],"challenge":[80],"lies":[81],"providing":[83],"effective":[84],"supervision":[85],"for":[86,101],"training":[87],"compressor.":[89],"end,":[92],"design":[94],"an":[95],"evolutionary":[96],"labeling":[97],"strategy":[98],"searches":[100],"subsets":[103],"minimizing":[104],"MLLM's":[106],"output":[107],"loss,":[108],"enforcing":[110],"semantic":[111,144],"diversity":[112],"through":[113],"vocabulary-based":[114],"grouping.":[116],"We":[117],"further":[118],"train":[119],"using":[122],"tailored":[124],"loss":[125,130],"function":[126],"combining":[127],"GHM":[129],"to":[131,142,184],"mitigate":[132],"class":[133],"difficulty":[135],"imbalance,":[136],"cosine":[139],"similarity":[140,167],"regularization":[141],"encourage":[143],"separation":[145],"between":[146],"retained":[147],"discarded":[149],"tokens.":[150],"Extensive":[151],"experiments":[152],"across":[153],"multiple":[154],"benchmarks":[156],"show":[157],"outperforms":[160],"existing":[161],"methods":[162],"based":[163],"attention":[165],"heuristics.":[168],"Notably,":[169],"it":[170],"retains":[171],"99.3%":[172],"original":[175],"accuracy":[176],"under":[177],"3x":[178],"delivers":[182],"up":[183],"1.6x":[185],"speedup":[186],"mobile":[188],"devices.":[189]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-22T00:00:00"}