{"id":"https://openalex.org/W7138882415","doi":"https://doi.org/10.48550/arxiv.2603.16737","title":"Retrieving Counterfactuals Improves Visual In-Context Learning","display_name":"Retrieving Counterfactuals Improves Visual In-Context Learning","publication_year":2026,"publication_date":"2026-03-17","ids":{"openalex":"https://openalex.org/W7138882415","doi":"https://doi.org/10.48550/arxiv.2603.16737"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.16737","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.16737","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.16737","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5047322230","display_name":"Guangzhi Xiong","orcid":"https://orcid.org/0000-0002-8049-5298"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Guangzhi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049731366","display_name":"Sanchit Sinha","orcid":"https://orcid.org/0000-0003-2650-0612"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sinha, Sanchit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130043557","display_name":"Zhenghao He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Zhenghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5129879741","display_name":"Aidong Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Aidong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.957099974155426,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.957099974155426,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.018699999898672104,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.005200000014156103,"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/spurious-relationship","display_name":"Spurious relationship","score":0.7224000096321106},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.6284999847412109},{"id":"https://openalex.org/keywords/counterfactual-conditional","display_name":"Counterfactual conditional","score":0.5123999714851379},{"id":"https://openalex.org/keywords/visual-reasoning","display_name":"Visual reasoning","score":0.48339998722076416},{"id":"https://openalex.org/keywords/limiting","display_name":"Limiting","score":0.41029998660087585},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.39259999990463257},{"id":"https://openalex.org/keywords/range","display_name":"Range (aeronautics)","score":0.3831000030040741},{"id":"https://openalex.org/keywords/causal-model","display_name":"Causal model","score":0.36649999022483826}],"concepts":[{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.7224000096321106},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.699400007724762},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.6284999847412109},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6054999828338623},{"id":"https://openalex.org/C71889745","wikidata":"https://www.wikidata.org/wiki/Q1783264","display_name":"Counterfactual conditional","level":3,"score":0.5123999714851379},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.48339998722076416},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.48330000042915344},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.41029998660087585},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.39259999990463257},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3831000030040741},{"id":"https://openalex.org/C11671645","wikidata":"https://www.wikidata.org/wiki/Q5054567","display_name":"Causal model","level":2,"score":0.36649999022483826},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.349700003862381},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.33489999175071716},{"id":"https://openalex.org/C115086926","wikidata":"https://www.wikidata.org/wiki/Q17004651","display_name":"Causal reasoning","level":3,"score":0.3124000132083893},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.30720001459121704},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2896000146865845},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.26829999685287476},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2563999891281128},{"id":"https://openalex.org/C166151441","wikidata":"https://www.wikidata.org/wiki/Q4923601","display_name":"Causation","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.16737","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.16737","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.16737","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.16737","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision-language":[0],"models":[1,177],"(VLMs)":[2],"have":[3],"achieved":[4],"impressive":[5],"performance":[6],"across":[7,150],"a":[8,34,90],"wide":[9],"range":[10],"of":[11,52],"multimodal":[12],"reasoning":[13],"tasks,":[14,43],"but":[15,44,69],"they":[16],"often":[17],"struggle":[18],"to":[19,39,41,66,115],"disentangle":[20],"fine-grained":[21],"visual":[22],"attributes":[23,123],"and":[24,75,124,130,134,168],"reason":[25,117],"about":[26,118],"underlying":[27],"causal":[28,120],"relationships.":[29],"In-context":[30],"learning":[31],"(ICL)":[32],"offers":[33],"promising":[35],"avenue":[36],"for":[37,85,181],"VLMs":[38,114],"adapt":[40],"new":[42],"its":[45],"effectiveness":[46],"critically":[47],"depends":[48],"on":[49,60,139,154],"the":[50,119],"selection":[51],"demonstration":[53,96],"examples.":[54],"Existing":[55],"retrieval-augmented":[56],"approaches":[57],"typically":[58],"rely":[59],"passive":[61],"similarity-based":[62],"retrieval,":[63],"which":[64],"tends":[65],"select":[67],"correlated":[68],"non-causal":[70],"examples,":[71,111,171],"amplifying":[72],"spurious":[73],"associations":[74],"limiting":[76],"model":[77],"robustness.":[78],"We":[79],"introduce":[80],"CIRCLES":[81,112,145,164],"(Composed":[82],"Image":[83],"Retrieval":[84],"Causal":[86],"Learning":[87],"Example":[88],"Selection),":[89],"novel":[91],"framework":[92],"that":[93,144],"actively":[94],"constructs":[95],"sets":[97],"by":[98],"retrieving":[99],"counterfactual-style":[100,110],"examples":[101],"through":[102],"targeted,":[103],"attribute-guided":[104],"composed":[105],"image":[106],"retrieval.":[107],"By":[108],"incorporating":[109],"enables":[113],"implicitly":[116],"relations":[121],"between":[122],"outcomes,":[125],"moving":[126],"beyond":[127],"superficial":[128],"correlations":[129],"fostering":[131],"more":[132,166],"robust":[133],"grounded":[135],"reasoning.":[136,183],"Comprehensive":[137],"experiments":[138],"four":[140],"diverse":[141,167],"datasets":[142],"demonstrate":[143],"consistently":[146],"outperforms":[147],"existing":[148],"methods":[149],"multiple":[151],"architectures,":[152],"especially":[153],"small-scale":[155],"models,":[156],"with":[157],"pronounced":[158],"gains":[159],"under":[160],"information":[161],"scarcity.":[162],"Furthermore,":[163],"retrieves":[165],"causally":[169],"informative":[170],"providing":[172],"qualitative":[173],"insights":[174],"into":[175],"how":[176],"leverage":[178],"in-context":[179],"demonstrations":[180],"improved":[182],"Our":[184],"code":[185],"is":[186],"available":[187],"at":[188],"https://github.com/gzxiong/CIRCLES.":[189]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-20T00:00:00"}