{"id":"https://openalex.org/W7153335819","doi":"https://doi.org/10.48550/arxiv.2604.07812","title":"HAWK: Head Importance-Aware Visual Token Pruning in Multimodal Models","display_name":"HAWK: Head Importance-Aware Visual Token Pruning in Multimodal Models","publication_year":2026,"publication_date":"2026-04-09","ids":{"openalex":"https://openalex.org/W7153335819","doi":"https://doi.org/10.48550/arxiv.2604.07812"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.07812","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07812","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.2604.07812","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133321834","display_name":"Qihui Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Qihui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133351669","display_name":"Tao Zhang","orcid":"https://orcid.org/0009-0000-6747-0267"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133354271","display_name":"Yuchen Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yuchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085950168","display_name":"Zijian Wen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wen, Zijian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133386610","display_name":"Mengjie Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Mengjie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045720734","display_name":"Shuangwu Chen","orcid":"https://orcid.org/0000-0003-2817-9738"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Shuangwu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133374960","display_name":"Xiaobin Tan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tan, Xiaobin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133371547","display_name":"Jian Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Jian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100638749","display_name":"Yang Liu","orcid":"https://orcid.org/0000-0002-4795-9169"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133390677","display_name":"Zhenhua Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Zhenhua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133360754","display_name":"Xianzhi Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Xianzhi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5070635874","display_name":"Yinfei Pan","orcid":"https://orcid.org/0000-0002-1830-1886"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pan, Yinfei","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":0.9240000247955322,"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":0.9240000247955322,"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":0.014100000262260437,"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":0.014000000432133675,"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":0.7272999882698059},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.6096000075340271},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4853000044822693},{"id":"https://openalex.org/keywords/visualization","display_name":"Visualization","score":0.4172999858856201},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.36500000953674316},{"id":"https://openalex.org/keywords/visual-perception","display_name":"Visual perception","score":0.34290000796318054},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.3206000030040741},{"id":"https://openalex.org/keywords/human-visual-system-model","display_name":"Human visual system model","score":0.30799999833106995}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8348000049591064},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7272999882698059},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.6096000075340271},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5875999927520752},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4853000044822693},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.4172999858856201},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.36500000953674316},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.34290000796318054},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3255000114440918},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3206000030040741},{"id":"https://openalex.org/C160086991","wikidata":"https://www.wikidata.org/wiki/Q5939193","display_name":"Human visual system model","level":3,"score":0.30799999833106995},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2955999970436096},{"id":"https://openalex.org/C2780312720","wikidata":"https://www.wikidata.org/wiki/Q5689100","display_name":"Head (geology)","level":2,"score":0.2944999933242798},{"id":"https://openalex.org/C178278151","wikidata":"https://www.wikidata.org/wiki/Q7936607","display_name":"Visual memory","level":3,"score":0.28940001130104065},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.27720001339912415},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.26420000195503235},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C207363949","wikidata":"https://www.wikidata.org/wiki/Q462915","display_name":"Visual space","level":3,"score":0.2540000081062317},{"id":"https://openalex.org/C2986089797","wikidata":"https://www.wikidata.org/wiki/Q6501338","display_name":"Visual attention","level":3,"score":0.25369998812675476},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2529999911785126},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2522999942302704}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.07812","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07812","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.2604.07812","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07812","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":[{"id":"https://metadata.un.org/sdg/8","display_name":"Decent work and economic growth","score":0.4250219762325287}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"In":[0,80],"multimodal":[1],"large":[2],"language":[3],"models":[4],"(MLLMs),":[5],"the":[6,13,36,57,97,108,171,178,189,198],"surge":[7],"of":[8,38,82,100,110,170,177,188],"visual":[9,44,58,70,78,91,104,123,130,179],"tokens":[10,131],"significantly":[11],"increases":[12],"inference":[14,40],"time":[15],"and":[16,72,118,142,191],"computational":[17],"overhead,":[18],"making":[19],"them":[20],"impractical":[21],"for":[22,34],"real-time":[23],"or":[24],"resource-constrained":[25],"applications.":[26],"Visual":[27],"token":[28,92,124],"pruning":[29,93,175],"is":[30,139,203],"a":[31,88],"promising":[32],"strategy":[33],"reducing":[35],"cost":[37],"MLLM":[39],"by":[41],"removing":[42,133],"redundant":[43,134],"tokens.":[45,112,180],"Existing":[46],"research":[47],"usually":[48],"assumes":[49],"that":[50,64,95,158],"all":[51],"attention":[52,101,120],"heads":[53,66,102],"contribute":[54],"equally":[55],"to":[56,106,121,147,165,186],"interpretation.":[59],"However,":[60],"our":[61],"study":[62],"reveals":[63],"different":[65],"may":[67],"capture":[68],"distinct":[69,75],"semantics":[71],"inherently":[73],"play":[74],"roles":[76],"in":[77,103],"processing.":[79],"light":[81],"this":[83],"observation,":[84],"we":[85],"propose":[86],"HAWK,":[87],"head":[89,115],"importance-aware":[90],"method":[94],"perceives":[96],"varying":[98],"importance":[99,116],"tasks":[105],"maximize":[107],"retention":[109],"crucial":[111],"By":[113],"leveraging":[114],"weights":[117],"text-guided":[119],"assess":[122],"significance,":[125],"HAWK":[126,138,159,167],"effectively":[127],"retains":[128,168],"task-relevant":[129],"while":[132],"ones.":[135],"The":[136,201],"proposed":[137],"entirely":[140],"training-free":[141],"can":[143],"be":[144],"seamlessly":[145],"applied":[146,164],"various":[148],"MLLMs.":[149],"Extensive":[150],"experiments":[151],"on":[152],"multiple":[153],"mainstream":[154],"vision-language":[155],"benchmarks":[156],"demonstrate":[157],"achieves":[160],"state-of-the-art":[161],"accuracy.":[162],"When":[163],"Qwen2.5-VL,":[166],"96.0%":[169],"original":[172,190],"accuracy":[173],"after":[174],"80.2%":[176],"Additionally,":[181],"it":[182],"reduces":[183],"end-to-end":[184],"latency":[185],"74.4%":[187],"further":[192],"decreases":[193],"GPU":[194],"memory":[195],"usage":[196],"across":[197],"tested":[199],"models.":[200],"code":[202],"available":[204],"at":[205],"https://github.com/peppery77/HAWK.git.":[206]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-11T00:00:00"}