{"id":"https://openalex.org/W7161797280","doi":"https://doi.org/10.48550/arxiv.2605.19929","title":"Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models","display_name":"Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161797280","doi":"https://doi.org/10.48550/arxiv.2605.19929"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.19929","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19929","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.2605.19929","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136559132","display_name":"Yi Zhong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136507357","display_name":"Haotong Qin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qin, Haotong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136567906","display_name":"Xindong Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xindong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136580801","display_name":"Lei Zhang","orcid":"https://orcid.org/0009-0000-2082-1162"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Lei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136570578","display_name":"Guolei Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Guolei","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":0.4584999978542328,"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":0.4584999978542328,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.21789999306201935,"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":0.06440000236034393,"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/quantization","display_name":"Quantization (signal processing)","score":0.776199996471405},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.6704999804496765},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.6412000060081482},{"id":"https://openalex.org/keywords/spurious-relationship","display_name":"Spurious relationship","score":0.41499999165534973},{"id":"https://openalex.org/keywords/source-code","display_name":"Source code","score":0.36550000309944153},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.3617999851703644},{"id":"https://openalex.org/keywords/anomaly-detection","display_name":"Anomaly detection","score":0.3589000105857849}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.776199996471405},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6859999895095825},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.6704999804496765},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.6412000060081482},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4702000021934509},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4442000091075897},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.41499999165534973},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.36550000309944153},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3617999851703644},{"id":"https://openalex.org/C739882","wikidata":"https://www.wikidata.org/wiki/Q3560506","display_name":"Anomaly detection","level":2,"score":0.3589000105857849},{"id":"https://openalex.org/C205606062","wikidata":"https://www.wikidata.org/wiki/Q5249645","display_name":"Decoupling (probability)","level":2,"score":0.35580000281333923},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.35280001163482666},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.351500004529953},{"id":"https://openalex.org/C127162648","wikidata":"https://www.wikidata.org/wiki/Q16858953","display_name":"Channel (broadcasting)","level":2,"score":0.3276999890804291},{"id":"https://openalex.org/C2780310539","wikidata":"https://www.wikidata.org/wiki/Q12547192","display_name":"Imperfect","level":2,"score":0.3165999948978424},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.31360000371932983},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.29789999127388},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.28439998626708984},{"id":"https://openalex.org/C141513077","wikidata":"https://www.wikidata.org/wiki/Q378542","display_name":"Independent and identically distributed random variables","level":3,"score":0.2696000039577484}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.19929","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19929","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.2605.19929","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19929","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":{"Low-bit":[0],"post-training":[1,76],"quantization":[2,128,150],"(PTQ)":[3],"is":[4,43,183],"a":[5,48,74,84],"pivotal":[6],"technique":[7],"for":[8,65,177],"deploying":[9,178],"Vision-Language":[10],"Models":[11],"(VLMs)":[12],"on":[13,132],"resource-constrained":[14],"devices.":[15],"However,":[16],"existing":[17,140],"PTQ":[18],"methods":[19],"often":[20],"degrade":[21],"VLMs'":[22],"accuracy":[23],"due":[24],"to":[25,124],"the":[26,105,166,174],"heterogeneous":[27],"activation":[28],"distributions":[29],"of":[30,51,162],"text":[31],"and":[32,57,156],"vision":[33],"modalities":[34],"during":[35],"quantization.":[36],"We":[37],"find":[38],"that":[39,92,118,136],"this":[40],"cross-modal":[41,107],"heterogeneity":[42],"distributed":[44],"unevenly":[45],"across":[46,142],"channels:":[47],"small":[49],"subset":[50],"channels":[52,64,98],"contains":[53],"most":[54],"modality-specific":[55,96],"outliers,":[56],"these":[58],"outliers":[59],"typically":[60],"reside":[61],"in":[62],"different":[63],"each":[66],"modality.":[67],"Motivated":[68],"by":[69],"this,":[70],"we":[71,110],"propose":[72],"SplitQ,":[73],"channel-Splitting-driven":[75],"Quantization":[77],"framework.":[78],"At":[79],"its":[80],"core,":[81],"SplitQ":[82,137,159],"introduces":[83],"novel":[85],"Modality-specific":[86],"Outlier":[87],"Channel":[88],"Decoupling":[89],"(MOCD)":[90],"module":[91,117],"effectively":[93],"isolates":[94],"salient":[95],"outlier":[97],"with":[99],"minimal":[100],"overhead.":[101],"To":[102],"further":[103],"address":[104],"remaining":[106],"distribution":[108],"discrepancies,":[109],"design":[111],"an":[112],"Adaptive":[113],"Cross-Modal":[114],"Calibration":[115],"(ACC)":[116],"employs":[119],"dual":[120],"lightweight":[121],"learnable":[122],"branches":[123],"dynamically":[125],"mitigate":[126],"modality-induced":[127],"errors.":[129],"Extensive":[130],"experiments":[131],"popular":[133,144],"VLMs":[134],"demonstrate":[135],"significantly":[138],"outperforms":[139],"approaches":[141],"6":[143],"multi-modal":[145],"datasets":[146],"under":[147,165],"all":[148],"evaluated":[149],"settings,":[151],"including":[152],"W4A8,":[153],"W4A4,":[154],"W3A3,":[155],"W3A2.":[157],"Notably,":[158],"preserves":[160],"93.5%":[161],"FP16":[163],"performance":[164],"challenging":[167],"W3A3":[168],"setting":[169],"(69.5":[170],"vs.":[171],"74.3),":[172],"pushing":[173],"efficiency":[175],"frontier":[176],"advanced":[179],"VLMs.":[180],"Our":[181],"code":[182],"available":[184],"at":[185],"https://github.com/EMVision-NK/SplitQ":[186]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-21T00:00:00"}