{"id":"https://openalex.org/W7155032784","doi":"https://doi.org/10.48550/arxiv.2604.17274","title":"Instinct vs. Reflection: Unifying Token and Verbalized Confidence in Multimodal Large Models","display_name":"Instinct vs. Reflection: Unifying Token and Verbalized Confidence in Multimodal Large Models","publication_year":2026,"publication_date":"2026-04-19","ids":{"openalex":"https://openalex.org/W7155032784","doi":"https://doi.org/10.48550/arxiv.2604.17274"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.17274","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17274","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.17274","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134198719","display_name":"Yunkai Dang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dang, Yunkai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134187976","display_name":"Yifan Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006110323","display_name":"Yizhu Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Yizhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134116059","display_name":"Anqi Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Anqi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134109112","display_name":"Wenbin Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Wenbin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134101023","display_name":"Yang Gao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"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/T10028","display_name":"Topic Modeling","score":0.36070001125335693,"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/T10028","display_name":"Topic Modeling","score":0.36070001125335693,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.24869999289512634,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.046799998730421066,"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/merge","display_name":"Merge (version control)","score":0.6150000095367432},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.5695000290870667},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4733999967575073},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.4641000032424927},{"id":"https://openalex.org/keywords/perception","display_name":"Perception","score":0.44769999384880066},{"id":"https://openalex.org/keywords/categorization","display_name":"Categorization","score":0.3882000148296356},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.3790999948978424}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6442000269889832},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.6150000095367432},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.5695000290870667},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5322999954223633},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4733999967575073},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.4641000032424927},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.44769999384880066},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4323999881744385},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.3882000148296356},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.3790999948978424},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.36399999260902405},{"id":"https://openalex.org/C44249647","wikidata":"https://www.wikidata.org/wiki/Q208498","display_name":"Confidence interval","level":2,"score":0.3628000020980835},{"id":"https://openalex.org/C3232509","wikidata":"https://www.wikidata.org/wiki/Q18237","display_name":"Instinct","level":2,"score":0.3472999930381775},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.27709999680519104},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2653000056743622},{"id":"https://openalex.org/C2834757","wikidata":"https://www.wikidata.org/wiki/Q4925424","display_name":"Monotone polygon","level":2,"score":0.2597000002861023},{"id":"https://openalex.org/C97508593","wikidata":"https://www.wikidata.org/wiki/Q1359069","display_name":"Contingency","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.25200000405311584}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.17274","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17274","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.17274","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.17274","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"Large":[1],"Language":[2],"Models":[3],"(MLLMs)":[4],"have":[5,30],"demonstrated":[6],"exceptional":[7],"capabilities":[8],"in":[9,21],"various":[10],"perception":[11],"and":[12,52,96,129,142,146],"reasoning":[13],"tasks.":[14],"Despite":[15],"this":[16,44,48,83],"success,":[17],"ensuring":[18],"their":[19],"reliability":[20],"practical":[22],"deployment":[23],"necessitates":[24],"robust":[25],"confidence":[26,60,89,140],"estimation.":[27,61],"Prior":[28],"works":[29],"predominantly":[31],"focused":[32],"on":[33,38,126],"text-only":[34],"LLMs,":[35],"often":[36],"relying":[37],"computationally":[39],"expensive":[40],"self-consistency":[41],"sampling.":[42],"In":[43],"paper,":[45],"we":[46,85],"extend":[47],"to":[49,92,99,110],"multimodal":[50],"settings":[51],"conduct":[53],"a":[54,65,87],"comprehensive":[55],"evaluation":[56],"of":[57],"MLLMs'":[58],"response":[59],"Our":[62],"analysis":[63],"reveals":[64],"significant":[66],"instinct-reflection":[67],"misalignment:":[68],"the":[69,119],"model's":[70],"implicit":[71],"token-level":[72],"support":[73],"frequently":[74],"diverges":[75],"from":[76],"its":[77],"verbal":[78],"self-assessment":[79],"confidence.":[80],"To":[81],"address":[82],"misalignment,":[84],"propose":[86],"monotone":[88],"fusion":[90],"framework":[91],"merge":[93],"dual-channel":[94],"signals":[95],"cross-channel":[97],"consistency":[98],"estimate":[100],"correctness.":[101],"Subsequently,":[102],"an":[103],"order-preserving":[104],"mean":[105],"alignment":[106],"step":[107],"is":[108],"applied":[109],"correct":[111],"global":[112],"bias,":[113],"which":[114],"improves":[115,143],"calibration":[116,145],"while":[117],"preserving":[118],"risk-coverage":[120],"trade-off":[121],"for":[122],"selective":[123],"prediction.":[124,148],"Experiments":[125],"diverse":[127],"open-source":[128],"closed-source":[130],"MLLMs":[131],"show":[132],"that":[133],"our":[134],"method":[135],"consistently":[136],"yields":[137],"more":[138],"reliable":[139],"estimates":[141],"both":[144],"failure":[147],"Code":[149],"will":[150],"be":[151],"available":[152],"at":[153],"https://github.com/Yunkaidang/Instinct-vs.-Reflection.":[154]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-22T00:00:00"}