{"id":"https://openalex.org/W7133541024","doi":"https://doi.org/10.48550/arxiv.2603.02556","title":"Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs","display_name":"Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs","publication_year":2026,"publication_date":"2026-03-03","ids":{"openalex":"https://openalex.org/W7133541024","doi":"https://doi.org/10.48550/arxiv.2603.02556"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.02556","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.02556","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.2603.02556","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128046535","display_name":"Zhiyu Pan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pan, Zhiyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087300058","display_name":"Yizheng Wu","orcid":"https://orcid.org/0000-0001-5335-4919"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yizheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082778608","display_name":"Jiashen Hua","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hua, Jiashen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128073221","display_name":"Junyi Feng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Junyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128051622","display_name":"Shaotian Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Shaotian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128076814","display_name":"Bing Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Bing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Cao, Zhiguo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cao, Zhiguo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5128102591","display_name":"Jieping Ye","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ye, Jieping","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":0.9373999834060669,"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":0.9373999834060669,"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/T11148","display_name":"Language, Metaphor, and Cognition","score":0.014000000432133675,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.0034000000450760126,"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/visual-reasoning","display_name":"Visual reasoning","score":0.8123999834060669},{"id":"https://openalex.org/keywords/semantic-reasoner","display_name":"Semantic reasoner","score":0.5981000065803528},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5753999948501587},{"id":"https://openalex.org/keywords/suite","display_name":"Suite","score":0.5443000197410583},{"id":"https://openalex.org/keywords/contrast","display_name":"Contrast (vision)","score":0.4984000027179718},{"id":"https://openalex.org/keywords/visualization","display_name":"Visualization","score":0.3596000075340271},{"id":"https://openalex.org/keywords/analytic-reasoning","display_name":"Analytic reasoning","score":0.3495999872684479},{"id":"https://openalex.org/keywords/reasoning-system","display_name":"Reasoning system","score":0.3384999930858612}],"concepts":[{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.8123999834060669},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7739999890327454},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6812000274658203},{"id":"https://openalex.org/C9616225","wikidata":"https://www.wikidata.org/wiki/Q3929429","display_name":"Semantic reasoner","level":2,"score":0.5981000065803528},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5753999948501587},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.5443000197410583},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.4984000027179718},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.47110000252723694},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3596000075340271},{"id":"https://openalex.org/C103057564","wikidata":"https://www.wikidata.org/wiki/Q4751139","display_name":"Analytic reasoning","level":3,"score":0.3495999872684479},{"id":"https://openalex.org/C89288958","wikidata":"https://www.wikidata.org/wiki/Q7301504","display_name":"Reasoning system","level":2,"score":0.3384999930858612},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.3379000127315521},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.311599999666214},{"id":"https://openalex.org/C37335422","wikidata":"https://www.wikidata.org/wiki/Q6888134","display_name":"Model-based reasoning","level":3,"score":0.3109999895095825},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.310699999332428},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.28999999165534973},{"id":"https://openalex.org/C193221554","wikidata":"https://www.wikidata.org/wiki/Q5153664","display_name":"Commonsense reasoning","level":2,"score":0.289900004863739},{"id":"https://openalex.org/C83725634","wikidata":"https://www.wikidata.org/wiki/Q7268699","display_name":"Qualitative reasoning","level":2,"score":0.2840000092983246},{"id":"https://openalex.org/C195344581","wikidata":"https://www.wikidata.org/wiki/Q2555318","display_name":"Automated reasoning","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.2621000111103058},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.25679999589920044},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2549999952316284},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.2529999911785126},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.02556","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.02556","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.2603.02556","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.02556","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":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.6785331964492798}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reasoning":[0],"has":[1],"emerged":[2],"as":[3],"a":[4,41,58,67,97,113,134],"key":[5,59],"capability":[6,15,148],"of":[7,116,149,183],"large":[8],"language":[9,37],"models.":[10],"In":[11],"linguistic":[12],"tasks,":[13],"this":[14,88],"can":[16,185],"be":[17,49],"enhanced":[18],"by":[19,87],"self-improving":[20,33,99,164],"techniques":[21],"that":[22,101,158,178],"refine":[23],"reasoning":[24,46,137,147,175],"paths":[25,47],"for":[26],"subsequent":[27],"finetuning.":[28,154],"However,":[29],"extending":[30],"these":[31],"language-based":[32],"approaches":[34,165],"to":[35,105,123,144],"vision":[36],"models":[38,169],"(VLMs)":[39],"presents":[40],"unique":[42],"challenge:~visual":[43],"hallucinations":[44,107],"in":[45,108],"cannot":[48],"effectively":[50],"verified":[51],"or":[52],"rectified.":[53],"Our":[54],"solution":[55],"starts":[56],"with":[57,66,76],"observation":[60],"about":[61],"visual":[62,82,103,136,174,189],"contrast:":[63],"when":[64],"presented":[65],"contrastive":[68,120,181],"VQA":[69,117],"pair,":[70],"i.e.,":[71],"two":[72],"visually":[73],"similar":[74],"images":[75],"synonymous":[77],"questions,":[78],"VLMs":[79,151,184],"identify":[80],"relevant":[81],"cues":[83],"more":[84],"precisely.":[85],"Motivated":[86],"observation,":[89],"we":[90,132],"propose":[91],"Visual":[92],"Contrastive":[93],"Self-Taught":[94],"Reasoner":[95],"(VC-STaR),":[96],"novel":[98],"framework":[100],"leverages":[102],"contrast":[104],"mitigate":[106],"model-generated":[109],"rationales.":[110],"We":[111],"collect":[112],"diverse":[114],"suite":[115],"datasets,":[118,176],"curate":[119],"pairs":[121],"according":[122],"multi-modal":[124],"similarity,":[125],"and":[126],"generate":[127],"rationales":[128],"using":[129],"VC-STaR.":[130],"Consequently,":[131],"obtain":[133],"new":[135],"dataset,":[138],"VisCoR-55K,":[139],"which":[140],"is":[141],"then":[142],"used":[143],"boost":[145],"the":[146,172,179],"various":[150],"through":[152],"supervised":[153],"Extensive":[155],"experiments":[156],"show":[157],"VC-STaR":[159],"not":[160],"only":[161],"outperforms":[162],"existing":[163],"but":[166],"also":[167],"surpasses":[168],"finetuned":[170],"on":[171],"SoTA":[173],"demonstrating":[177],"inherent":[180],"ability":[182],"bootstrap":[186],"their":[187],"own":[188],"reasoning.":[190],"Project":[191],"at:":[192],"https://github.com/zhiyupan42/VC-STaR.":[193]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-05T00:00:00"}