{"id":"https://openalex.org/W4415345064","doi":"https://doi.org/10.1145/3774521.3774581","title":"Strongly Connected Components Are All You Need: Graph-Theoretic Interpretability and Optimization for Vision Transformers","display_name":"Strongly Connected Components Are All You Need: Graph-Theoretic Interpretability and Optimization for Vision Transformers","publication_year":2025,"publication_date":"2025-12-17","ids":{"openalex":"https://openalex.org/W4415345064","doi":"https://doi.org/10.1145/3774521.3774581"},"language":null,"primary_location":{"id":"doi:10.1145/3774521.3774581","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3774521.3774581","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Sixteen Indian Conference on Computer Vision, Graphics and Image Processing","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3774521.3774581","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119994619","display_name":"Devansh Garg","orcid":null},"institutions":[{"id":"https://openalex.org/I9579091","display_name":"Indian Institute of Technology Mandi","ror":"https://ror.org/05r9r2f34","country_code":"IN","type":"education","lineage":["https://openalex.org/I9579091"]}],"countries":["IN"],"is_corresponding":true,"raw_author_name":"Devansh Garg","raw_affiliation_strings":["Indian Institute of Technology Mandi, Mandi, Himachal Pradesh, India","Indian Institute of Technology, Mandi"],"raw_orcid":"https://orcid.org/0009-0000-8210-2526","affiliations":[{"raw_affiliation_string":"Indian Institute of Technology Mandi, Mandi, Himachal Pradesh, India","institution_ids":["https://openalex.org/I9579091"]},{"raw_affiliation_string":"Indian Institute of Technology, Mandi","institution_ids":["https://openalex.org/I9579091"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5119994619"],"corresponding_institution_ids":["https://openalex.org/I9579091"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.31292406,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12292","display_name":"Graph Theory and Algorithms","score":0.9725000262260437,"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/T12292","display_name":"Graph Theory and Algorithms","score":0.9725000262260437,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.9017000198364258,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11273","display_name":"Advanced Graph Neural Networks","score":0.9010999798774719,"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/interpretability","display_name":"Interpretability","score":0.8345000147819519},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6464999914169312},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.595300018787384},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.5076000094413757},{"id":"https://openalex.org/keywords/through-the-lens-metering","display_name":"Through-the-lens metering","score":0.47920000553131104},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.36640000343322754},{"id":"https://openalex.org/keywords/feature-learning","display_name":"Feature learning","score":0.3612000048160553},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.3562000095844269}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.8345000147819519},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.71670001745224},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6464999914169312},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.595300018787384},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5706999897956848},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.5076000094413757},{"id":"https://openalex.org/C43091099","wikidata":"https://www.wikidata.org/wiki/Q1067788","display_name":"Through-the-lens metering","level":3,"score":0.47920000553131104},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4569000005722046},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.36640000343322754},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.3612000048160553},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.3562000095844269},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3465999960899353},{"id":"https://openalex.org/C2987255567","wikidata":"https://www.wikidata.org/wiki/Q33002955","display_name":"Knowledge graph","level":2,"score":0.3237999975681305},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.3077000081539154},{"id":"https://openalex.org/C34736171","wikidata":"https://www.wikidata.org/wiki/Q918333","display_name":"Preprocessor","level":2,"score":0.3043999969959259},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2985000014305115},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.2906999886035919},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.26489999890327454},{"id":"https://openalex.org/C5339829","wikidata":"https://www.wikidata.org/wiki/Q1425977","display_name":"Machine vision","level":2,"score":0.26089999079704285},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C106937863","wikidata":"https://www.wikidata.org/wiki/Q7236518","display_name":"Power graph analysis","level":3,"score":0.2500999867916107}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3774521.3774581","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3774521.3774581","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Sixteen Indian Conference on Computer Vision, Graphics and Image Processing","raw_type":"proceedings-article"},{"id":"doi:10.21203/rs.3.rs-7877883/v1","is_oa":true,"landing_page_url":"https://doi.org/10.21203/rs.3.rs-7877883/v1","pdf_url":null,"source":{"id":"https://openalex.org/S4306525896","display_name":"Research Square","issn_l":null,"issn":null,"is_oa":false,"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":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":"Research Square","raw_type":"posted-content"}],"best_oa_location":{"id":"doi:10.1145/3774521.3774581","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3774521.3774581","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Sixteen Indian Conference on Computer Vision, Graphics and Image Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W2118382442","https://openalex.org/W2194775991","https://openalex.org/W2907492528","https://openalex.org/W2919115771","https://openalex.org/W2963341956","https://openalex.org/W2970726176","https://openalex.org/W3176196997"],"related_works":[],"abstract_inverted_index":{"<title>Abstract</title>":[0],"Vision":[1,70],"Transformers":[2],"have":[3],"revolutionized":[4],"computer":[5],"vision":[6,217],"with":[7,117,159],"unprecedented":[8],"performance,":[9],"yet":[10],"their":[11],"attention":[12,41,72,77,95,153],"mechanisms":[13],"remain":[14],"largely":[15],"opaque,":[16],"hindering":[17],"both":[18],"trust":[19],"and":[20,82,108,183,210],"optimization":[21,141],"in":[22,30],"critical":[23],"applications":[24],"[1,":[25],"2].":[26],"Despite":[27],"significant":[28],"advances":[29],"transformer":[31,208],"interpretability,":[32,197],"existing":[33],"approaches":[34],"provide":[35],"limited":[36],"structural":[37,92,175],"insights":[38],"into":[39],"how":[40],"patterns":[42,93],"encode":[43],"visual":[44,125],"information,":[45],"creating":[46],"a":[47,63,127,139,204],"fundamental":[48],"barrier":[49],"to":[50,89,122,180],"understanding":[51,131,222],"these":[52],"powerful":[53,205],"models":[54,104,172,218],"[3,":[55],"4].":[56],"We":[57],"address":[58],"this":[59,198],"challenge":[60],"by":[61],"introducing":[62],"novel,":[64],"comprehensive":[65],"graph-theoretic":[66],"framework":[67],"for":[68,130,164,207,214],"analyzing":[69],"Transformer":[71],"mechanisms.":[73],"Our":[74,167],"approach":[75],"treats":[76],"matrices":[78],"as":[79,203],"directed":[80],"graphs":[81],"applies":[83],"Tarjan\u2019s":[84],"Strongly":[85,114],"Connected":[86,115],"Components":[87,116],"algorithm":[88],"reveal":[90],"hidden":[91],"within":[94],"flows":[96],"[5].":[97],"Through":[98],"systematic":[99],"evaluation":[100],"of":[101],"22":[102],"pre-trained":[103],"across":[105],"ViT,":[106],"DeiT,":[107],"CLIP":[109,184],"architectures,":[110],"we":[111],"discover":[112],"that":[113,170],"meaningful":[118,146],"connectivity":[119,147],"directly":[120],"correspond":[121],"semantically":[123],"coherent":[124],"features\u2014providing":[126],"mathematical":[128],"foundation":[129],"attention-based":[132],"feature":[133],"learning":[134],"[6].":[135],"This":[136],"insight":[137],"enables":[138],"novel":[140],"strategy:":[142],"layers":[143],"containing":[144],"minimal":[145],"can":[148],"be":[149],"selectively":[150],"ablated":[151],"through":[152],"linearization,":[154],"achieving":[155],"16-30%":[156],"inference":[157],"speedup":[158],"only":[160],"0-7%":[161],"accuracy":[162],"loss":[163],"large":[165],"models.":[166],"findings":[168],"demonstrate":[169],"ViT":[171],"exhibit":[173],"superior":[174],"organization":[176],"(92.6%":[177],"effectiveness)":[178],"compared":[179],"DeiT":[181],"(51.1%)":[182],"(57.2%)":[185],"variants,":[186],"while":[187],"patch32":[188],"configurations":[189],"achieve":[190],"optimal":[191],"performance":[192],"tradeoffs":[193],"[7,":[194],"8].":[195],"Beyond":[196],"work":[199],"establishes":[200],"graph":[201],"theory":[202],"lens":[206],"analysis":[209],"provides":[211],"practical":[212],"tools":[213],"deploying":[215],"efficient":[216],"without":[219],"sacrificing":[220],"semantic":[221],"[9].":[223]},"counts_by_year":[],"updated_date":"2026-08-02T14:50:37.381335","created_date":"2025-10-21T00:00:00"}