{"id":"https://openalex.org/W4406857603","doi":"https://doi.org/10.1109/tmm.2025.3535405","title":"Prune and Merge: Efficient Token Compression for Vision Transformer With Spatial Information Preserved","display_name":"Prune and Merge: Efficient Token Compression for Vision Transformer With Spatial Information Preserved","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4406857603","doi":"https://doi.org/10.1109/tmm.2025.3535405"},"language":"en","primary_location":{"id":"doi:10.1109/tmm.2025.3535405","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tmm.2025.3535405","pdf_url":null,"source":{"id":"https://openalex.org/S137030581","display_name":"IEEE Transactions on Multimedia","issn_l":"1520-9210","issn":["1520-9210","1941-0077"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Multimedia","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5003709886","display_name":"Junzhu Mao","orcid":"https://orcid.org/0009-0005-0831-9272"},"institutions":[{"id":"https://openalex.org/I36399199","display_name":"Nanjing University of Science and Technology","ror":"https://ror.org/00xp9wg62","country_code":"CN","type":"education","lineage":["https://openalex.org/I36399199"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Junzhu Mao","raw_affiliation_strings":["School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China"],"raw_orcid":"https://orcid.org/0009-0005-0831-9272","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China","institution_ids":["https://openalex.org/I36399199"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014424245","display_name":"Yang Shen","orcid":"https://orcid.org/0000-0002-6344-9951"},"institutions":[{"id":"https://openalex.org/I36399199","display_name":"Nanjing University of Science and Technology","ror":"https://ror.org/00xp9wg62","country_code":"CN","type":"education","lineage":["https://openalex.org/I36399199"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang Shen","raw_affiliation_strings":["School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0002-6344-9951","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China","institution_ids":["https://openalex.org/I36399199"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039703932","display_name":"Jinyang Guo","orcid":"https://orcid.org/0000-0003-1956-3367"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jinyang Guo","raw_affiliation_strings":["State Key Laboratory of Software Development Environment, Institute of Artificial Intelligence, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1956-3367","affiliations":[{"raw_affiliation_string":"State Key Laboratory of Software Development Environment, Institute of Artificial Intelligence, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027545344","display_name":"Yazhou Yao","orcid":"https://orcid.org/0000-0002-0337-9410"},"institutions":[{"id":"https://openalex.org/I36399199","display_name":"Nanjing University of Science and Technology","ror":"https://ror.org/00xp9wg62","country_code":"CN","type":"education","lineage":["https://openalex.org/I36399199"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yazhou Yao","raw_affiliation_strings":["School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0002-0337-9410","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China","institution_ids":["https://openalex.org/I36399199"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024965898","display_name":"Xian\u2010Sheng Hua","orcid":"https://orcid.org/0000-0002-8232-5049"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiansheng Hua","raw_affiliation_strings":["Terminus Group, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-8232-5049","affiliations":[{"raw_affiliation_string":"Terminus Group, Beijing, China","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5052993469","display_name":"Heng Tao Shen","orcid":"https://orcid.org/0000-0002-2999-2088"},"institutions":[{"id":"https://openalex.org/I150229711","display_name":"University of Electronic Science and Technology of China","ror":"https://ror.org/04qr3zq92","country_code":"CN","type":"education","lineage":["https://openalex.org/I150229711"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hengtao Shen","raw_affiliation_strings":["School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, China"],"raw_orcid":"https://orcid.org/0000-0002-2999-2088","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, China","institution_ids":["https://openalex.org/I150229711"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":8.0792,"has_fulltext":false,"cited_by_count":9,"citation_normalized_percentile":{"value":0.97436378,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":"27","issue":null,"first_page":"4670","last_page":"4683"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12923","display_name":"Digital Image Processing Techniques","score":0.9904000163078308,"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/T12923","display_name":"Digital Image Processing Techniques","score":0.9904000163078308,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.9868000149726868,"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/T11017","display_name":"Chaos-based Image/Signal Encryption","score":0.9846000075340271,"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/computer-science","display_name":"Computer science","score":0.8398488163948059},{"id":"https://openalex.org/keywords/merge","display_name":"Merge (version control)","score":0.6358422636985779},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5802623629570007},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.44575855135917664},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4195009469985962},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3585919141769409},{"id":"https://openalex.org/keywords/computer-vision","display_name":"Computer vision","score":0.3386523127555847},{"id":"https://openalex.org/keywords/information-retrieval","display_name":"Information retrieval","score":0.18885430693626404},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.16542306542396545}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8398488163948059},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.6358422636985779},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5802623629570007},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.44575855135917664},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4195009469985962},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3585919141769409},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3386523127555847},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.18885430693626404},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.16542306542396545},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tmm.2025.3535405","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tmm.2025.3535405","pdf_url":null,"source":{"id":"https://openalex.org/S137030581","display_name":"IEEE Transactions on Multimedia","issn_l":"1520-9210","issn":["1520-9210","1941-0077"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Multimedia","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1482089936","display_name":null,"funder_award_id":"BK20240080","funder_id":"https://openalex.org/F4320322769","funder_display_name":"Natural Science Foundation of Jiangsu Province"},{"id":"https://openalex.org/G5800900978","display_name":null,"funder_award_id":"62472222","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320322769","display_name":"Natural Science Foundation of Jiangsu Province","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":79,"referenced_works":["https://openalex.org/W2108598243","https://openalex.org/W2507296351","https://openalex.org/W2962858109","https://openalex.org/W2963563276","https://openalex.org/W2965862774","https://openalex.org/W2980088508","https://openalex.org/W2981851019","https://openalex.org/W3007007518","https://openalex.org/W3096609285","https://openalex.org/W3121523901","https://openalex.org/W3131500599","https://openalex.org/W3138516171","https://openalex.org/W3170841864","https://openalex.org/W3188427387","https://openalex.org/W3214466318","https://openalex.org/W3216553632","https://openalex.org/W4206281850","https://openalex.org/W4214588794","https://openalex.org/W4214614183","https://openalex.org/W4214633470","https://openalex.org/W4214893857","https://openalex.org/W4295899642","https://openalex.org/W4304080557","https://openalex.org/W4312233661","https://openalex.org/W4312340826","https://openalex.org/W4312349930","https://openalex.org/W4312769131","https://openalex.org/W4312910119","https://openalex.org/W4313007769","https://openalex.org/W4313156423","https://openalex.org/W4321021073","https://openalex.org/W4322730938","https://openalex.org/W4323065857","https://openalex.org/W4362654014","https://openalex.org/W4362683541","https://openalex.org/W4365446402","https://openalex.org/W4376467320","https://openalex.org/W4384159680","https://openalex.org/W4385572090","https://openalex.org/W4386066311","https://openalex.org/W4386076556","https://openalex.org/W4390190626","https://openalex.org/W4390874575","https://openalex.org/W4391021637","https://openalex.org/W4394593136","https://openalex.org/W4402667905","https://openalex.org/W4402716428","https://openalex.org/W4402780389","https://openalex.org/W4403021086","https://openalex.org/W4404004660","https://openalex.org/W4411245291","https://openalex.org/W6677103964","https://openalex.org/W6726983090","https://openalex.org/W6732814185","https://openalex.org/W6734169003","https://openalex.org/W6739917289","https://openalex.org/W6757817989","https://openalex.org/W6762718338","https://openalex.org/W6762945437","https://openalex.org/W6763835125","https://openalex.org/W6768080748","https://openalex.org/W6768429542","https://openalex.org/W6775432837","https://openalex.org/W6779436764","https://openalex.org/W6780100354","https://openalex.org/W6780482815","https://openalex.org/W6784094891","https://openalex.org/W6784333009","https://openalex.org/W6788135285","https://openalex.org/W6793979934","https://openalex.org/W6796494063","https://openalex.org/W6796580215","https://openalex.org/W6797478244","https://openalex.org/W6798160016","https://openalex.org/W6810818354","https://openalex.org/W6838719481","https://openalex.org/W6839091850","https://openalex.org/W6845607375","https://openalex.org/W6846577953"],"related_works":["https://openalex.org/W4388335561","https://openalex.org/W2970530566","https://openalex.org/W4288261899","https://openalex.org/W4307309205","https://openalex.org/W2967478618","https://openalex.org/W4385009901","https://openalex.org/W4385572700","https://openalex.org/W2997152889","https://openalex.org/W4234886518","https://openalex.org/W4304700937"],"abstract_inverted_index":{"Token":[0],"compression":[1,30,106,124],"is":[2],"essential":[3],"for":[4,99],"reducing":[5],"the":[6,73,93,97,115,129,137,183],"computational":[7],"and":[8,27,34,41,56,59,71,104,120,132,176,193],"memory":[9],"requirements":[10],"of":[11,75,118,139,186,191],"transformer":[12,45],"models,":[13],"enabling":[14,72],"their":[15],"deployment":[16],"in":[17,167,189],"resource-constrained":[18],"environments.":[19],"In":[20],"this":[21],"work,":[22],"we":[23,63,79,157,181],"propose":[24],"an":[25],"efficient":[26],"hardware-compatible":[28],"token":[29,39,50,89],"method":[31],"called":[32],"Prune":[33],"Merge.":[35],"Our":[36],"approach":[37,188],"integrates":[38],"pruning":[40],"merging":[42],"operations":[43],"within":[44],"models":[46,175],"to":[47,113,150],"achieve":[48,158],"layer-wise":[49],"compression.":[51],"By":[52],"introducing":[53],"trainable":[54],"merge":[55,65],"reconstruct":[57],"matrices":[58],"utilizing":[60],"shortcut":[61],"connections,":[62],"efficiently":[64],"tokens":[66,119],"while":[67],"preserving":[68],"important":[69],"information":[70,112],"restoration":[74],"pruned":[76],"tokens.":[77],"Additionally,":[78],"introduce":[80],"a":[81,159,164],"novel":[82],"gradient-weighted":[83],"attention":[84],"scoring":[85],"mechanism":[86],"that":[87],"computes":[88],"importance":[90],"scores":[91],"during":[92,102],"training":[94],"phase,":[95],"eliminating":[96],"need":[98],"separate":[100],"computations":[101],"inference":[103],"enhancing":[105],"efficiency.":[107],"We":[108],"also":[109],"leverage":[110],"gradient":[111],"capture":[114],"global":[116],"impact":[117],"automatically":[121],"identify":[122],"optimal":[123],"structures.":[125],"Extensive":[126],"experiments":[127],"on":[128,155,169],"ImageNet-1":[130],"k":[131],"ADE20":[133],"K":[134],"datasets":[135],"validate":[136],"effectiveness":[138],"our":[140,187],"approach,":[141],"achieving":[142],"significant":[143],"speed-ups":[144],"with":[145,162,178],"minimal":[146],"accuracy":[147,168],"degradation":[148],"compared":[149],"state-of-the-art":[151],"methods.":[152],"For":[153],"instance,":[154],"DeiT-Small,":[156],"1.64\u00d7":[160],"speed-up":[161],"only":[163],"0.2%":[165],"drop":[166],"ImageNet-1k.":[170],"Moreover,":[171],"by":[172],"compressing":[173],"segmenter":[174],"comparing":[177],"existing":[179],"methods,":[180],"demonstrate":[182],"superior":[184],"performance":[185],"terms":[190],"efficiency":[192],"effectiveness.":[194]},"counts_by_year":[{"year":2026,"cited_by_count":3},{"year":2025,"cited_by_count":6}],"updated_date":"2026-06-19T17:40:00.097472","created_date":"2025-10-10T00:00:00"}