{"id":"https://openalex.org/W7214766169","doi":"https://doi.org/10.48550/arxiv.2609.34867","title":"P4Q: Co-designing Token Pruning and Quantization for Vision-Language Model Acceleration","display_name":"P4Q: Co-designing Token Pruning and Quantization for Vision-Language Model Acceleration","publication_year":2026,"publication_date":"2026-09-28","ids":{"openalex":"https://openalex.org/W7214766169","doi":"https://doi.org/10.48550/arxiv.2609.34867"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.34867","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34867","pdf_url":"https://arxiv.org/pdf/2609.34867","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.34867","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5031323156","display_name":"Haizhao Jing","orcid":"https://orcid.org/0000-0002-4147-331X","observed_orcids":["https://orcid.org/0000-0002-4147-331X"]},"institutions":[{"id":"https://openalex.org/I17145004","display_name":"Northwestern Polytechnical University","ror":"https://ror.org/01y0j0j86","country_code":"CN","type":"education","lineage":["https://openalex.org/I17145004"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jing, Haizhao","raw_affiliation_strings":["Northwest Polytechnical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northwest Polytechnical University","institution_ids":["https://openalex.org/I17145004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125548745","display_name":"Zhenhao Shang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I17145004","display_name":"Northwestern Polytechnical University","ror":"https://ror.org/01y0j0j86","country_code":"CN","type":"education","lineage":["https://openalex.org/I17145004"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shang, Zhenhao","raw_affiliation_strings":["Northwest Polytechnical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northwest Polytechnical University","institution_ids":["https://openalex.org/I17145004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153422960","display_name":"Haokui Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I17145004","display_name":"Northwestern Polytechnical University","ror":"https://ror.org/01y0j0j86","country_code":"CN","type":"education","lineage":["https://openalex.org/I17145004"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Haokui","raw_affiliation_strings":["Northwest Polytechnical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northwest Polytechnical University","institution_ids":["https://openalex.org/I17145004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084417688","display_name":"Rong Xiao","orcid":"https://orcid.org/0000-0002-6408-9724","observed_orcids":["https://orcid.org/0000-0002-6408-9724"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, Rong","raw_affiliation_strings":["Intellifusion"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intellifusion","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153459139","display_name":"Peng Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I17145004","display_name":"Northwestern Polytechnical University","ror":"https://ror.org/01y0j0j86","country_code":"CN","type":"education","lineage":["https://openalex.org/I17145004"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Peng","raw_affiliation_strings":["Northwest Polytechnical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northwest Polytechnical University","institution_ids":["https://openalex.org/I17145004"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9878000020980835,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9878000020980835,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.008299999870359898,"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":0.0005000000237487257,"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/vision-language-models","display_name":"vision-language models","score":0.8619999885559082},{"id":"https://openalex.org/keywords/visual-token-pruning","display_name":"visual token pruning","score":0.949999988079071},{"id":"https://openalex.org/keywords/post-training-quantization","display_name":"post-training quantization","score":0.8899999856948853},{"id":"https://openalex.org/keywords/low-bit-quantization","display_name":"low-bit quantization","score":0.8399999737739563},{"id":"https://openalex.org/keywords/co-design","display_name":"co-design","score":0.5649999976158142},{"id":"https://openalex.org/keywords/model-compression","display_name":"model compression","score":0.6919999718666077},{"id":"https://openalex.org/keywords/inference-efficiency","display_name":"inference efficiency","score":0.746999979019165},{"id":"https://openalex.org/keywords/llava-next","display_name":"LLaVA-NeXT","score":0.8240000009536743}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8421000242233276},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7483999729156494},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6646000146865845},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6119999885559082},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5403000116348267},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.4828000068664551},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.44029998779296875},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.41929998993873596},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38960000872612},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.3605000078678131},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2863999903202057},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.28439998626708984},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.28139999508857727},{"id":"https://openalex.org/C136625980","wikidata":"https://www.wikidata.org/wiki/Q663208","display_name":"Rounding","level":2,"score":0.27630001306533813},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.27459999918937683},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.2736999988555908},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26969999074935913},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.2669999897480011},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.2524999976158142}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.34867","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34867","pdf_url":"https://arxiv.org/pdf/2609.34867","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.34867","doi":"https://doi.org/10.48550/arxiv.2609.34867","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.34867","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.34867","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34867","pdf_url":"https://arxiv.org/pdf/2609.34867","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214766169.pdf","grobid_xml":"https://content.openalex.org/works/W7214766169.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W4385473806","https://openalex.org/W4389523971","https://openalex.org/W4402671893","https://openalex.org/W7081567423","https://openalex.org/W7128485811","https://openalex.org/W7139051607","https://openalex.org/W7150989117","https://openalex.org/W7155072443","https://openalex.org/W7169880607"],"related_works":["https://openalex.org/W7155072443","https://openalex.org/W7150989117","https://openalex.org/W7206184220","https://openalex.org/W7081567423","https://openalex.org/W7214003627","https://openalex.org/W4417450014","https://openalex.org/W7138352833","https://openalex.org/W7077077575","https://openalex.org/W4415795654","https://openalex.org/W7127919000"],"abstract_inverted_index":{"Vision":[0],"language":[1],"models":[2],"have":[3],"achieved":[4],"strong":[5],"performance":[6],"across":[7,214],"a":[8,75,95,141,193],"wide":[9],"range":[10],"of":[11,110,212],"multimodal":[12],"applications,":[13],"yet":[14],"their":[15],"substantial":[16,183],"computational":[17],"and":[18,27,40,62,72,85,117,226],"memory":[19],"costs":[20],"hinder":[21],"efficient":[22,89],"deployment.":[23,175],"Visual":[24],"token":[25,83,98],"pruning":[26,84,153],"post-training":[28],"quantization":[29,87,107,143,227],"reduce":[30],"inference":[31,184,210],"overhead":[32],"along":[33],"two":[34,179],"complementary":[35],"dimensions,":[36],"namely":[37],"sequence":[38],"length":[39],"numerical":[41],"precision.":[42],"Existing":[43],"workflows":[44],"typically":[45],"optimize":[46],"these":[47,70,125,178],"techniques":[48],"independently":[49,198],"or":[50],"apply":[51],"them":[52],"sequentially.":[53],"Their":[54],"distinct":[55,216],"optimization":[56],"objectives":[57],"leave":[58],"critical":[59],"interactions":[60],"unaddressed":[61],"constrain":[63],"the":[64,102,111,115,130,148,156,160,165,169],"achievable":[65],"compression":[66,225],"performance.":[67],"We":[68],"revisit":[69],"designs":[71],"present":[73],"P4Q,":[74],"practical":[76],"co-design":[77],"framework":[78],"that":[79],"jointly":[80],"optimizes":[81],"visual":[82,97,119],"low-bit":[86,136],"for":[88],"VLM":[90],"inference.":[91],"First,":[92],"P4Q":[93,139,181,205],"introduces":[94,140],"quantization-aware":[96],"selection":[99,150],"strategy":[100,151],"before":[101],"LLM.":[103],"It":[104,146],"applies":[105],"fake":[106],"to":[108,154],"copies":[109],"features":[112],"produced":[113],"by":[114],"projector":[116],"selects":[118],"tokens":[120],"using":[121],"statistics":[122],"computed":[123],"from":[124],"fake-quantized":[126],"features,":[127],"thereby":[128,163],"conditioning":[129],"selector's":[131],"feature-based":[132],"decisions":[133],"on":[134,159,203],"simulated":[135],"perturbations.":[137],"Second,":[138],"pruning-aware":[142],"calibration":[144,166],"strategy.":[145],"uses":[147],"same":[149],"as":[152],"calibrate":[155],"quantized":[157],"model":[158],"retained-token":[161],"distribution,":[162],"aligning":[164],"process":[167],"with":[168],"pruned":[170],"execution":[171],"path":[172],"used":[173],"during":[174],"By":[176],"coupling":[177],"components,":[180],"achieves":[182,206],"speedups":[185],"while":[186,219],"maintaining":[187],"comparable":[188],"task":[189],"performance,":[190],"resulting":[191],"in":[192],"better":[194],"efficiency-accuracy":[195],"trade-off":[196],"than":[197,223],"optimized":[199],"pipelines.":[200],"For":[201],"instance,":[202],"LLaVA-NeXT,":[204],"an":[207],"average":[208],"end-to-end":[209],"speedup":[211],"2.8x":[213],"eight":[215],"test":[217],"sets,":[218],"retaining":[220],"higher":[221],"accuracy":[222],"prior":[224],"methods.":[228]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-30T00:00:00"}