{"id":"https://openalex.org/W7220909751","doi":"https://doi.org/10.1109/cvprf71909.2026.00316","title":"Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP","display_name":"Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP","publication_year":2026,"publication_date":"2026-06-01","ids":{"openalex":"https://openalex.org/W7220909751","doi":"https://doi.org/10.1109/cvprf71909.2026.00316"},"language":"en","primary_location":{"id":"doi:10.1109/cvprf71909.2026.00316","doi":"https://doi.org/10.1109/cvprf71909.2026.00316","is_oa":true,"landing_page_url":"https://doi.org/10.1109/cvprf71909.2026.00316","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":"2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition- FINDINGS Track (CVPRF)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1109/cvprf71909.2026.00316","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133174359","display_name":"Yusung Ro","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Yusung Ro","raw_affiliation_strings":["KAIST"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"KAIST","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035688668","display_name":"Jaehyun Choi","orcid":"https://orcid.org/0000-0002-9183-761X","observed_orcids":["https://orcid.org/0000-0002-9183-761X"]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jaehyun Choi","raw_affiliation_strings":["KAIST"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"KAIST","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5159059732","display_name":"Junmo Kim","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Junmo Kim","raw_affiliation_strings":["KAIST"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"KAIST","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I157485424"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.68035544,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3252","last_page":"3261"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9955999851226807,"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"}},"topics":[{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9955999851226807,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.003000000026077032,"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":0.0006000000284984708,"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/sparse-autoencoders","display_name":"sparse autoencoders","score":0.9990000128746033},{"id":"https://openalex.org/keywords/clip","display_name":"CLIP","score":0.8399999737739563},{"id":"https://openalex.org/keywords/information-scope","display_name":"information scope","score":0.8130000233650208},{"id":"https://openalex.org/keywords/interpretability","display_name":"interpretability","score":0.8299999833106995},{"id":"https://openalex.org/keywords/model-confidence","display_name":"model confidence","score":0.37700000405311584}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6395999789237976},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5389999747276306},{"id":"https://openalex.org/C2778012447","wikidata":"https://www.wikidata.org/wiki/Q1034415","display_name":"Scope (computer science)","level":2,"score":0.5322999954223633},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3474000096321106},{"id":"https://openalex.org/C180198813","wikidata":"https://www.wikidata.org/wiki/Q121182","display_name":"Information system","level":2,"score":0.2732999920845032},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.25540000200271606},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.2547000050544739},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.25270000100135803},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.24709999561309814}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1109/cvprf71909.2026.00316","doi":"https://doi.org/10.1109/cvprf71909.2026.00316","is_oa":true,"landing_page_url":"https://doi.org/10.1109/cvprf71909.2026.00316","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":"2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition- FINDINGS Track (CVPRF)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2604.05724","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2604.05724","pdf_url":"https://arxiv.org/pdf/2604.05724","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.2604.05724","doi":"https://doi.org/10.48550/arxiv.2604.05724","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.05724","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":"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"}],"best_oa_location":{"id":"doi:10.1109/cvprf71909.2026.00316","doi":"https://doi.org/10.1109/cvprf71909.2026.00316","is_oa":true,"landing_page_url":"https://doi.org/10.1109/cvprf71909.2026.00316","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":"2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition- FINDINGS Track (CVPRF)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G5834210620","display_name":null,"funder_award_id":"2025-02283048","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G6183154644","display_name":null,"funder_award_id":"RS-2025-02283048","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"}],"funders":[{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7220909751.pdf","grobid_xml":"https://content.openalex.org/works/W7220909751.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W2039089492","https://openalex.org/W2105464873","https://openalex.org/W2108598243","https://openalex.org/W2143668817","https://openalex.org/W2737258237","https://openalex.org/W3107364113","https://openalex.org/W4224058739","https://openalex.org/W4403843009","https://openalex.org/W4404987606","https://openalex.org/W4410348438","https://openalex.org/W4416035733","https://openalex.org/W4416036800","https://openalex.org/W4417097950","https://openalex.org/W7131106712"],"related_works":["https://openalex.org/W7196967119","https://openalex.org/W4415798266","https://openalex.org/W7213573899","https://openalex.org/W4415795561","https://openalex.org/W7145927952","https://openalex.org/W7166876200","https://openalex.org/W4402684222","https://openalex.org/W4416033768","https://openalex.org/W4404783416","https://openalex.org/W4412886966"],"abstract_inverted_index":{"Sparse":[0],"Autoencoders":[1],"(SAEs)":[2],"have":[3],"emerged":[4],"as":[5,34,132],"a":[6,35,79,133,143],"powerful":[7],"tool":[8],"for":[9,137],"interpreting":[10],"the":[11,24,91],"internal":[12],"representations":[13,140],"of":[14,27,38,114,147],"CLIP":[15,139],"vision":[16],"encoders,":[17],"yet":[18],"existing":[19],"analyses":[20],"largely":[21],"focus":[22],"on":[23,122],"semantic":[25],"meaning":[26],"individual":[28],"features.":[29,108,149],"We":[30,59],"introduce":[31],"information":[32,116,130],"scope":[33,101,107,131],"complementary":[36],"dimension":[37],"interpretability":[39],"that":[40,61,112],"characterizes":[41],"how":[42],"broadly":[43],"an":[44],"SAE":[45,63],"feature":[46],"aggregates":[47],"visual":[48],"evidence,":[49],"ranging":[50],"from":[51,103],"localized,":[52],"patchspecific":[53],"cues":[54],"to":[55],"global,":[56],"image-level":[57],"signals.":[58],"observe":[60],"some":[62],"features":[64,102,113],"respond":[65],"consistently":[66],"across":[67],"spatial":[68],"perturbations,":[69],"while":[70],"others":[71],"shift":[72],"unpredictably":[73],"with":[74],"minor":[75],"input":[76],"changes,":[77],"indicating":[78],"fundamental":[80],"distinction":[81],"in":[82],"their":[83],"underlying":[84],"scope.":[85],"To":[86],"quantify":[87],"this,":[88],"we":[89],"propose":[90],"Contextual":[92],"Dependency":[93],"Score":[94],"(CDS),":[95],"which":[96],"separates":[97],"positionally":[98,104],"stable":[99],"local":[100],"variant":[105],"global":[106],"Our":[109],"experiments":[110],"show":[111],"different":[115,120],"scopes":[117],"exert":[118],"systematically":[119],"influences":[121],"CLIP's":[123],"predictions":[124],"and":[125,141],"confidence.":[126],"These":[127],"findings":[128],"establish":[129],"critical":[134],"new":[135],"axis":[136],"understanding":[138],"provide":[142],"deeper":[144],"diagnostic":[145],"view":[146],"SAE-derived":[148]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-10-09T00:00:00"}