{"id":"https://openalex.org/W7125796887","doi":"https://doi.org/10.1007/s42979-025-04707-6","title":"Where Do Tokens Go? Understanding Pruning Behaviors in STEP at High Resolutions","display_name":"Where Do Tokens Go? Understanding Pruning Behaviors in STEP at High Resolutions","publication_year":2026,"publication_date":"2026-01-27","ids":{"openalex":"https://openalex.org/W7125796887","doi":"https://doi.org/10.1007/s42979-025-04707-6"},"language":"en","primary_location":{"id":"doi:10.1007/s42979-025-04707-6","is_oa":true,"landing_page_url":"https://doi.org/10.1007/s42979-025-04707-6","pdf_url":"https://link.springer.com/content/pdf/10.1007/s42979-025-04707-6.pdf","source":{"id":"https://openalex.org/S4210174798","display_name":"SN Computer Science","issn_l":"2661-8907","issn":["2661-8907","2662-995X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319965","host_organization_name":"Springer Nature","host_organization_lineage":["https://openalex.org/P4310319965"],"host_organization_lineage_names":["Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SN Computer Science","raw_type":"journal-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://link.springer.com/content/pdf/10.1007/s42979-025-04707-6.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124040813","display_name":"Michal Szczepanski","orcid":null},"institutions":[{"id":"https://openalex.org/I2738703131","display_name":"Commissariat \u00e0 l'\u00c9nergie Atomique et aux \u00c9nergies Alternatives","ror":"https://ror.org/00jjx8s55","country_code":"FR","type":"government","lineage":["https://openalex.org/I2738703131"]},{"id":"https://openalex.org/I277688954","display_name":"Universit\u00e9 Paris-Saclay","ror":"https://ror.org/03xjwb503","country_code":"FR","type":"education","lineage":["https://openalex.org/I277688954"]},{"id":"https://openalex.org/I4210085861","display_name":"Laboratoire d'Int\u00e9gration des Syst\u00e8mes et des Technologies","ror":"https://ror.org/000dbcc61","country_code":"FR","type":"government","lineage":["https://openalex.org/I2738703131","https://openalex.org/I2738703131","https://openalex.org/I277688954","https://openalex.org/I4210085861","https://openalex.org/I4210117989"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Michal Szczepanski","raw_affiliation_strings":["Universit\u00e9 Paris-Saclay, CEA, List, 91120, Palaiseau, France"],"raw_orcid":"https://orcid.org/0009-0000-9061-4396","affiliations":[{"raw_affiliation_string":"Universit\u00e9 Paris-Saclay, CEA, List, 91120, Palaiseau, France","institution_ids":["https://openalex.org/I277688954","https://openalex.org/I2738703131","https://openalex.org/I4210085861"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123952914","display_name":"Martyna Poreba","orcid":null},"institutions":[{"id":"https://openalex.org/I2738703131","display_name":"Commissariat \u00e0 l'\u00c9nergie Atomique et aux \u00c9nergies Alternatives","ror":"https://ror.org/00jjx8s55","country_code":"FR","type":"government","lineage":["https://openalex.org/I2738703131"]},{"id":"https://openalex.org/I277688954","display_name":"Universit\u00e9 Paris-Saclay","ror":"https://ror.org/03xjwb503","country_code":"FR","type":"education","lineage":["https://openalex.org/I277688954"]},{"id":"https://openalex.org/I4210085861","display_name":"Laboratoire d'Int\u00e9gration des Syst\u00e8mes et des Technologies","ror":"https://ror.org/000dbcc61","country_code":"FR","type":"government","lineage":["https://openalex.org/I2738703131","https://openalex.org/I2738703131","https://openalex.org/I277688954","https://openalex.org/I4210085861","https://openalex.org/I4210117989"]}],"countries":["FR"],"is_corresponding":true,"raw_author_name":"Martyna Poreba","raw_affiliation_strings":["Universit\u00e9 Paris-Saclay, CEA, List, 91120, Palaiseau, France"],"raw_orcid":"https://orcid.org/0000-0002-5102-7735","affiliations":[{"raw_affiliation_string":"Universit\u00e9 Paris-Saclay, CEA, List, 91120, Palaiseau, France","institution_ids":["https://openalex.org/I277688954","https://openalex.org/I2738703131","https://openalex.org/I4210085861"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5123952712","display_name":"Karim Haroun","orcid":null},"institutions":[{"id":"https://openalex.org/I1294671590","display_name":"Centre National de la Recherche Scientifique","ror":"https://ror.org/02feahw73","country_code":"FR","type":"government","lineage":["https://openalex.org/I1294671590"]},{"id":"https://openalex.org/I201841394","display_name":"Universit\u00e9 C\u00f4te d'Azur","ror":"https://ror.org/019tgvf94","country_code":"FR","type":"education","lineage":["https://openalex.org/I201841394"]},{"id":"https://openalex.org/I4210106479","display_name":"Laboratoire d'Informatique, Signaux et Syst\u00e8mes de Sophia Antipolis","ror":"https://ror.org/01215r597","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I1294671590","https://openalex.org/I1326498283","https://openalex.org/I201841394","https://openalex.org/I4210106479","https://openalex.org/I4210159245"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Karim Haroun","raw_affiliation_strings":["I3S, Universit\u00e9 C\u00f4te d\u2019Azur, CNRS, 06900, Sophia Antipolis, France"],"raw_orcid":"https://orcid.org/0009-0000-6972-6019","affiliations":[{"raw_affiliation_string":"I3S, Universit\u00e9 C\u00f4te d\u2019Azur, CNRS, 06900, Sophia Antipolis, France","institution_ids":["https://openalex.org/I201841394","https://openalex.org/I1294671590","https://openalex.org/I4210106479"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5123952914"],"corresponding_institution_ids":["https://openalex.org/I2738703131","https://openalex.org/I277688954","https://openalex.org/I4210085861"],"apc_list":{"value":2290,"currency":"EUR","value_usd":2890},"apc_paid":{"value":2290,"currency":"EUR","value_usd":2890},"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.11761038,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"7","issue":"2","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":0.7113000154495239,"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":0.7113000154495239,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.06639999896287918,"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"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.048700001090765,"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/encoder","display_name":"Encoder","score":0.7013000249862671},{"id":"https://openalex.org/keywords/computational-complexity-theory","display_name":"Computational complexity theory","score":0.6855999827384949},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.6593000292778015},{"id":"https://openalex.org/keywords/segmentation","display_name":"Segmentation","score":0.6504999995231628},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5584999918937683},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.5501000285148621},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.5467000007629395},{"id":"https://openalex.org/keywords/image-segmentation","display_name":"Image segmentation","score":0.37610000371932983}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7908999919891357},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.7013000249862671},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.6855999827384949},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6593000292778015},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.6504999995231628},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5584999918937683},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.5501000285148621},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5467000007629395},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5442000031471252},{"id":"https://openalex.org/C124504099","wikidata":"https://www.wikidata.org/wiki/Q56933","display_name":"Image segmentation","level":3,"score":0.37610000371932983},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3479999899864197},{"id":"https://openalex.org/C160633673","wikidata":"https://www.wikidata.org/wiki/Q355198","display_name":"Pixel","level":2,"score":0.32690000534057617},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.32030001282691956},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3107999861240387},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3059000074863434},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.30480000376701355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2964000105857849},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.29120001196861267},{"id":"https://openalex.org/C110384440","wikidata":"https://www.wikidata.org/wiki/Q1143270","display_name":"Upsampling","level":3,"score":0.2815000116825104},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2655999958515167},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.2630000114440918},{"id":"https://openalex.org/C2781039887","wikidata":"https://www.wikidata.org/wiki/Q1391724","display_name":"Factor (programming language)","level":2,"score":0.262800008058548},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.2590999901294708}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1007/s42979-025-04707-6","is_oa":true,"landing_page_url":"https://doi.org/10.1007/s42979-025-04707-6","pdf_url":"https://link.springer.com/content/pdf/10.1007/s42979-025-04707-6.pdf","source":{"id":"https://openalex.org/S4210174798","display_name":"SN Computer Science","issn_l":"2661-8907","issn":["2661-8907","2662-995X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319965","host_organization_name":"Springer Nature","host_organization_lineage":["https://openalex.org/P4310319965"],"host_organization_lineage_names":["Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SN Computer Science","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:2509.14165","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2509.14165","pdf_url":"https://arxiv.org/pdf/2509.14165","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"doi:10.1007/s42979-025-04707-6","is_oa":true,"landing_page_url":"https://doi.org/10.1007/s42979-025-04707-6","pdf_url":"https://link.springer.com/content/pdf/10.1007/s42979-025-04707-6.pdf","source":{"id":"https://openalex.org/S4210174798","display_name":"SN Computer Science","issn_l":"2661-8907","issn":["2661-8907","2662-995X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319965","host_organization_name":"Springer Nature","host_organization_lineage":["https://openalex.org/P4310319965"],"host_organization_lineage_names":["Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SN Computer Science","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320324252","display_name":"Universit\u00e9 Paris-Saclay","ror":"https://ror.org/03xjwb503"},{"id":"https://openalex.org/F4320334322","display_name":"HORIZON EUROPE Framework Programme","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7125796887.pdf","grobid_xml":"https://content.openalex.org/works/W7125796887.grobid-xml"},"referenced_works_count":49,"referenced_works":["https://openalex.org/W2117539524","https://openalex.org/W2340897893","https://openalex.org/W2561196672","https://openalex.org/W2737258237","https://openalex.org/W3131500599","https://openalex.org/W3138516171","https://openalex.org/W3170841864","https://openalex.org/W3180037928","https://openalex.org/W3188427387","https://openalex.org/W4214893857","https://openalex.org/W4221144813","https://openalex.org/W4285601701","https://openalex.org/W4312290555","https://openalex.org/W4312530435","https://openalex.org/W4312671791","https://openalex.org/W4312849330","https://openalex.org/W4312872526","https://openalex.org/W4312910119","https://openalex.org/W4313069943","https://openalex.org/W4319300012","https://openalex.org/W4362500802","https://openalex.org/W4382239283","https://openalex.org/W4385768233","https://openalex.org/W4385805202","https://openalex.org/W4386075588","https://openalex.org/W4390190613","https://openalex.org/W4390873017","https://openalex.org/W4390873032","https://openalex.org/W4390873361","https://openalex.org/W4394593136","https://openalex.org/W4394597024","https://openalex.org/W4394625744","https://openalex.org/W4400526093","https://openalex.org/W4402716414","https://openalex.org/W4402716428","https://openalex.org/W4402753804","https://openalex.org/W4402916260","https://openalex.org/W4403548495","https://openalex.org/W4403908286","https://openalex.org/W4404433224","https://openalex.org/W4405838811","https://openalex.org/W4406631984","https://openalex.org/W4406857603","https://openalex.org/W4408111774","https://openalex.org/W4409189413","https://openalex.org/W4409262834","https://openalex.org/W4409262904","https://openalex.org/W4413147663","https://openalex.org/W4415797474"],"related_works":[],"abstract_inverted_index":{"Abstract":[0],"Vision":[1],"Transformers":[2],"(ViTs)":[3],"achieve":[4],"state-of-the-art":[5],"performance":[6],"in":[7,125,133,156,164],"semantic":[8,83],"segmentation":[9,84],"but":[10],"are":[11],"hindered":[12],"by":[13,106],"high":[14],"computational":[15,75,126,157],"and":[16,26,37,92,128,159,191],"memory":[17],"costs.":[18],"To":[19],"address":[20],"this,":[21],"we":[22],"propose":[23],"STEP":[24,51,144,180],"(SuperToken":[25],"Early-Pruning),":[27],"a":[28,54,107,121,129,152,160,168],"hybrid":[29],"token-reduction":[30],"framework":[31,145],"that":[32,59,94],"combines":[33],"dynamic":[34],"patch":[35],"merging":[36,62],"token":[38,101],"pruning":[39],"to":[40,70,89,112,151,183],"enhance":[41],"efficiency":[42],"without":[43],"significantly":[44],"compromising":[45],"accuracy.":[46],"At":[47],"the":[48,100,113,139,142,178,195],"core":[49],"of":[50,109,172,185],"is":[52,97],"dCTS,":[53],"lightweight":[55],"CNN-based":[56],"policy":[57],"network":[58],"enables":[60],"flexible":[61],"into":[63],"superpatches.":[64],"Encoder":[65],"blocks":[66],"integrate":[67],"also":[68],"early-exits":[69],"remove":[71],"high-confident":[72],"supertokens,":[73],"lowering":[74],"load.":[76],"We":[77],"evaluate":[78],"our":[79],"method":[80],"on":[81],"high-resolution":[82],"benchmarks,":[85],"including":[86],"images":[87],"up":[88,150,182],"$$1024\\times1024$$":[90],",":[91],"show":[93],"when":[95,135],"dCTS":[96],"applied":[98],"alone,":[99],"count":[102],"can":[103,187],"be":[104,188],"reduced":[105],"factor":[108],"2.5":[110],"compared":[111],"standard":[114],"$$16\\times$$":[115],"pixel":[116],"patching":[117],"scheme.":[118],"This":[119],"yields":[120],"$$2.6\\times":[122],"$$":[123,131,154,162],"reduction":[124,155],"cost":[127],"$$3.4\\times":[130],"increase":[132],"throughput":[134],"using":[136],"ViT-Large":[137],"as":[138],"backbone.":[140],"Applying":[141],"full":[143],"further":[146],"improves":[147],"efficiency,":[148],"reaching":[149,194],"$$4\\times":[153],"complexity":[158],"$$1.7\\times":[161],"gain":[163],"inference":[165],"speed,":[166],"with":[167],"maximum":[169],"accuracy":[170],"drop":[171],"no":[173],"more":[174],"than":[175],"2.0%.":[176],"With":[177],"proposed":[179],"configurations,":[181],"40%":[184],"tokens":[186],"confidently":[189],"predicted":[190],"halted":[192],"before":[193],"final":[196],"encoder":[197],"layer.":[198]},"counts_by_year":[],"updated_date":"2026-05-23T08:51:43.019350","created_date":"2026-01-28T00:00:00"}