{"id":"https://openalex.org/W7162442089","doi":"https://doi.org/10.48550/arxiv.2605.24019","title":"MGVQ: Synergizing Multi-dimensional Sensitivity-Aware and Gradient-Hessian Fusion for Vector Quantization","display_name":"MGVQ: Synergizing Multi-dimensional Sensitivity-Aware and Gradient-Hessian Fusion for Vector Quantization","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7162442089","doi":"https://doi.org/10.48550/arxiv.2605.24019"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.24019","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.24019","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.24019","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137014251","display_name":"Zhong Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137021826","display_name":"Zukang Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Zukang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137021142","display_name":"Xing Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Xing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137027523","display_name":"Dawei Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Dawei","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/T10901","display_name":"Advanced Data Compression Techniques","score":0.675599992275238,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.675599992275238,"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/T11019","display_name":"Image Enhancement Techniques","score":0.036400001496076584,"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":0.03359999880194664,"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/quantization","display_name":"Quantization (signal processing)","score":0.7318999767303467},{"id":"https://openalex.org/keywords/codebook","display_name":"Codebook","score":0.6674000024795532},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.6362000107765198},{"id":"https://openalex.org/keywords/sensitivity","display_name":"Sensitivity (control systems)","score":0.38280001282691956},{"id":"https://openalex.org/keywords/linde\u2013buzo\u2013gray-algorithm","display_name":"Linde\u2013Buzo\u2013Gray algorithm","score":0.37470000982284546},{"id":"https://openalex.org/keywords/residual","display_name":"Residual","score":0.337799996137619},{"id":"https://openalex.org/keywords/quadratic-equation","display_name":"Quadratic equation","score":0.33640000224113464},{"id":"https://openalex.org/keywords/minification","display_name":"Minification","score":0.32820001244544983}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7318999767303467},{"id":"https://openalex.org/C127759330","wikidata":"https://www.wikidata.org/wiki/Q637416","display_name":"Codebook","level":2,"score":0.6674000024795532},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.6362000107765198},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6197999715805054},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5845999717712402},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.38280001282691956},{"id":"https://openalex.org/C93372532","wikidata":"https://www.wikidata.org/wiki/Q6552455","display_name":"Linde\u2013Buzo\u2013Gray algorithm","level":3,"score":0.37470000982284546},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.337799996137619},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.33640000224113464},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.32820001244544983},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.31459999084472656},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.29989999532699585},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.2904999852180481},{"id":"https://openalex.org/C40567965","wikidata":"https://www.wikidata.org/wiki/Q1820283","display_name":"Learning vector quantization","level":3,"score":0.2879999876022339},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.27900001406669617},{"id":"https://openalex.org/C177225278","wikidata":"https://www.wikidata.org/wiki/Q192674","display_name":"Factoring","level":2,"score":0.2777999937534332},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C165021410","wikidata":"https://www.wikidata.org/wiki/Q55564","display_name":"Lossy compression","level":2,"score":0.2687999904155731},{"id":"https://openalex.org/C25432639","wikidata":"https://www.wikidata.org/wiki/Q3711790","display_name":"Weight distribution","level":2,"score":0.25529998540878296},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C117251300","wikidata":"https://www.wikidata.org/wiki/Q1849855","display_name":"Parametric statistics","level":2,"score":0.2533999979496002}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.24019","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.24019","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.24019","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.24019","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision-Language":[0],"Models":[1],"(VLMs)":[2],"achieve":[3],"outstanding":[4],"performance,":[5],"yet":[6],"their":[7],"huge":[8],"model":[9,23,35,53],"size":[10],"severely":[11],"hinders":[12],"deployment":[13,227],"on":[14,175,211],"edge":[15],"devices":[16],"with":[17],"limited":[18],"resources.":[19],"As":[20],"an":[21],"efficient":[22,219],"compression":[24],"technique,":[25],"vector":[26,113],"quantization":[27,114,132,190,197],"(VQ)":[28],"excels":[29],"in":[30,40,232],"ultra-low-bit":[31,220],"representation,":[32],"which":[33],"maps":[34],"weights":[36],"to":[37,44,58,138,168],"discrete":[38],"codewords":[39],"a":[41,81,111,201],"compact":[42],"codebook":[43],"cut":[45],"memory":[46],"consumption":[47],"and":[48,73,103,120,144,162,165,181,218],"transmission":[49],"overhead":[50],"while":[51],"preserving":[52],"capability.":[54],"Direct":[55],"VQ":[56],"application":[57],"VLMs":[59,177],"still":[60],"has":[61],"two":[62,126],"core":[63,127],"limitations.":[64],"First,":[65],"cross-modality":[66],"weight":[67,95],"distribution":[68],"differences":[69],"brought":[70],"by":[71,80],"visual":[72],"textual":[74],"inputs":[75],"cannot":[76],"be":[77],"well":[78],"fitted":[79],"single":[82],"unified":[83],"codebook.":[84],"Second,":[85],"current":[86],"second-order":[87,153],"error":[88,154,160],"compensation":[89,105,155],"ignores":[90],"first-order":[91,157],"gradient":[92,101],"information,":[93],"causing":[94],"deviation":[96],"from":[97],"pre-trained":[98],"optimal":[99],"states,":[100],"drift":[102],"biased":[104],"results.":[106],"This":[107],"work":[108],"proposes":[109],"MGVQ,":[110],"novel":[112],"framework":[115],"integrating":[116],"multi-dimensional":[117],"sensitivity":[118,140,146],"perception":[119],"gradient-Hessian":[121],"fusion.":[122],"It":[123],"consists":[124],"of":[125,186,205,228],"modules:":[128],"sensitivity-guided":[129],"structured":[130],"mixed-precision":[131],"dynamically":[133],"assigns":[134],"different":[135],"bit-widths":[136],"according":[137],"channel":[139],"via":[141],"combined":[142],"global":[143],"local":[145],"analysis":[147],"for":[148],"refined":[149],"resource":[150],"allocation;":[151],"gradient-aware":[152],"embeds":[156],"gradients":[158],"into":[159],"correction,":[161],"adopts":[163],"Kronecker":[164],"Block-LDL":[166],"decomposition":[167],"ensure":[169],"low":[170],"computational":[171],"cost.":[172],"Extensive":[173],"experiments":[174],"mainstream":[176],"including":[178],"LLaVA-onevision,":[179],"InternVL2":[180],"Qwen2-VL":[182],"verify":[183],"the":[184,225],"effectiveness":[185],"MGVQ.":[187],"In":[188],"2-bit":[189],"settings,":[191],"MGVQ":[192],"surpasses":[193],"existing":[194],"advanced":[195],"post-training":[196],"methods":[198],"significantly,":[199],"achieving":[200],"maximum":[202],"accuracy":[203],"improvement":[204],"4.9":[206],"points":[207],"(71.4%":[208],"vs":[209],"67.0%":[210],"InternVL2-26B).":[212],"The":[213],"proposed":[214],"method":[215],"realizes":[216],"stable":[217],"VLM":[221],"quantization,":[222],"greatly":[223],"promoting":[224],"practical":[226],"multimodal":[229],"large":[230],"models":[231],"resource-limited":[233],"environments.":[234]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-27T00:00:00"}