{"id":"https://openalex.org/W7202006226","doi":null,"title":"Support Operation Factorization: Compositional Readout of Frozen Vision Encoders under Controlled Interventions","display_name":"Support Operation Factorization: Compositional Readout of Frozen Vision Encoders under Controlled Interventions","publication_year":2026,"publication_date":"2026-08-06","ids":{"openalex":"https://openalex.org/W7202006226"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.06174","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06174","pdf_url":"https://arxiv.org/pdf/2608.06174","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"},"volume":null,"type":"preprint","indexed_in":["arxiv"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.06174","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5159781317","display_name":"Zhongyao Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Zhongyao","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087818121","display_name":"Wanli Ouyang","orcid":"https://orcid.org/0000-0002-9163-2761","observed_orcids":["https://orcid.org/0000-0002-9163-2761"]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Ouyang, Wanli","raw_affiliation_strings":["MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041810936","display_name":"Taoyong Cui","orcid":"https://orcid.org/0000-0002-5532-6636","observed_orcids":["https://orcid.org/0000-0002-5532-6636"]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Cui, Taoyong","raw_affiliation_strings":["Department of Computer Science and Engineering , The Chinese University of Hong Kong","MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]},{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5146126050","display_name":"Pheng Ann Heng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Heng, Pheng Ann","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.49950000643730164,"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":0.49950000643730164,"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":0.2011999934911728,"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":0.1543000042438507,"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/compositional-analysis","display_name":"compositional analysis","score":0.7519999742507935},{"id":"https://openalex.org/keywords/dinov3","display_name":"DINOv3","score":0.6399999856948853},{"id":"https://openalex.org/keywords/siglip","display_name":"SigLIP","score":0.8870000243186951},{"id":"https://openalex.org/keywords/mujoco","display_name":"MuJoCo","score":0.5230000019073486},{"id":"https://openalex.org/keywords/coco","display_name":"COCO","score":0.7710000276565552},{"id":"https://openalex.org/keywords/vision-language-models","display_name":"vision-language models","score":0.3230000138282776},{"id":"https://openalex.org/keywords/interpretability","display_name":"interpretability","score":0.42899999022483826}],"concepts":[{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.6313999891281128},{"id":"https://openalex.org/C128107574","wikidata":"https://www.wikidata.org/wiki/Q182003","display_name":"Injective function","level":2,"score":0.6029999852180481},{"id":"https://openalex.org/C187691185","wikidata":"https://www.wikidata.org/wiki/Q2020720","display_name":"Grid","level":2,"score":0.5855000019073486},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5575000047683716},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.5508999824523926},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.45590001344680786},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.39899998903274536},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.3824000060558319},{"id":"https://openalex.org/C108154423","wikidata":"https://www.wikidata.org/wiki/Q1469792","display_name":"Salience (neuroscience)","level":2,"score":0.38179999589920044},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.37380000948905945},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3684000074863434},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.36489999294281006},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.36079999804496765},{"id":"https://openalex.org/C84111939","wikidata":"https://www.wikidata.org/wiki/Q5125465","display_name":"Clamping","level":2,"score":0.352400004863739},{"id":"https://openalex.org/C132459708","wikidata":"https://www.wikidata.org/wiki/Q744069","display_name":"Extrapolation","level":2,"score":0.3391000032424927},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3255999982357025},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.31540000438690186},{"id":"https://openalex.org/C138187205","wikidata":"https://www.wikidata.org/wiki/Q131251","display_name":"Tangent","level":2,"score":0.3075000047683716},{"id":"https://openalex.org/C183763347","wikidata":"https://www.wikidata.org/wiki/Q120976","display_name":"Factorial","level":2,"score":0.2937999963760376},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.2833000123500824},{"id":"https://openalex.org/C16469947","wikidata":"https://www.wikidata.org/wiki/Q2400745","display_name":"Fractional factorial design","level":3,"score":0.27810001373291016},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.27390000224113464},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2736000120639801},{"id":"https://openalex.org/C160633673","wikidata":"https://www.wikidata.org/wiki/Q355198","display_name":"Pixel","level":2,"score":0.2718999981880188},{"id":"https://openalex.org/C44154836","wikidata":"https://www.wikidata.org/wiki/Q45045","display_name":"Simulation","level":1,"score":0.27070000767707825},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C32022120","wikidata":"https://www.wikidata.org/wiki/Q797225","display_name":"Interference (communication)","level":3,"score":0.2667999863624573},{"id":"https://openalex.org/C136085584","wikidata":"https://www.wikidata.org/wiki/Q910289","display_name":"Overlay","level":2,"score":0.26510000228881836},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2572999894618988},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2542000114917755},{"id":"https://openalex.org/C199639397","wikidata":"https://www.wikidata.org/wiki/Q1788588","display_name":"Engineering drawing","level":1,"score":0.2542000114917755}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:arXiv.org:2608.06174","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06174","pdf_url":"https://arxiv.org/pdf/2608.06174","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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.06174","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06174","pdf_url":"https://arxiv.org/pdf/2608.06174","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"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7202006226.pdf","grobid_xml":"https://content.openalex.org/works/W7202006226.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7140290522","https://openalex.org/W3151884877","https://openalex.org/W7124716190","https://openalex.org/W3144340139","https://openalex.org/W7161641893","https://openalex.org/W7220377769","https://openalex.org/W3024805531","https://openalex.org/W7201833970","https://openalex.org/W7213905682","https://openalex.org/W7170090349"],"abstract_inverted_index":{"Compositional":[0],"analysis":[1],"of":[2],"frozen":[3,100],"vision":[4],"encoders":[5],"should":[6],"determine":[7],"both":[8],"what":[9],"changed":[10],"and":[11,23,52,63,88,112,126,148,171,182],"where":[12],"it":[13],"changed.":[14],"Standard":[15],"factor":[16],"probes":[17],"score":[18],"these":[19],"axes":[20],"separately,":[21],"however,":[22],"can":[24,92],"reward":[25],"multiple":[26],"operations":[27],"that":[28,56,73,90],"reuse":[29],"the":[30,78,84,119,134,155],"same":[31],"predicted":[32],"slot.":[33],"We":[34,40],"call":[35],"this":[36],"failure":[37,198],"operation":[38,50,66],"laundering.":[39],"introduce":[41],"an":[42],"injectively":[43],"aligned":[44],"leave-one-cell-out":[45],"protocol":[46],"over":[47,144],"support":[48,61],"x":[49],"grids":[51],"SO-OPF,":[53],"a":[54,64,145,163,196],"readout":[55,181],"factors":[57],"cell":[58,97],"energy":[59],"into":[60],"salience":[62],"competitive":[65],"posterior.":[67],"This":[68],"formulation":[69],"separates":[70],"two":[71,189],"questions":[72],"aggregate":[74],"scores":[75],"conflate:":[76],"whether":[77,89],"carrier":[79,136,147],"composes":[80],"held-out":[81,186],"bindings":[82,187],"when":[83],"grid":[85,91],"is":[86,167],"known,":[87],"be":[93],"recovered":[94],"from":[95,121,140,206],"flat":[96,122,207],"labels.":[98,208],"With":[99],"DINOv3":[101,170],"features,":[102],"known":[103],"factorial":[104],"assignment":[105,120],"reaches":[106,124],"0.874":[107],"injective":[108,183],"accuracy":[109,139,166],"on":[110,114,132,188],"Shapes3D-Extended":[111],"0.799":[113],"globally":[115],"image-disjoint":[116],"COCO;":[117],"learning":[118],"labels":[123],"0.769":[125],"0.762,":[127],"respectively.":[128],"Under":[129],"matched-axis-aware":[130],"supervision":[131],"Shapes3D,":[133],"factored":[135,180],"improves":[137],"learned-assignment":[138,165],"0.653":[141],"to":[142],"0.841":[143],"dense":[146],"eliminates":[149],"its":[150],"laundering":[151],"gap.":[152],"SigLIP2":[153],"replicates":[154],"COCO":[156],"separation.":[157],"A":[158],"rebuilt":[159],"MuJoCo":[160],"substrate":[161],"exposes":[162],"boundary:":[164],"0.569":[168],"with":[169,173,175],"0.484":[172],"SigLIP2,":[174],"substantial":[176],"slot":[177],"collapse.":[178],"Thus":[179],"evaluation":[184],"recover":[185],"substrates":[190],"while":[191],"exposing,":[192],"rather":[193],"than":[194],"hiding,":[195],"renderer-specific":[197],"boundary;":[199],"they":[200],"do":[201],"not":[202],"establish":[203],"universal":[204],"recovery":[205]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-09T00:00:00"}