{"id":"https://openalex.org/W7155403592","doi":"https://doi.org/10.48550/arxiv.2604.19974","title":"Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders","display_name":"Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders","publication_year":2026,"publication_date":"2026-04-21","ids":{"openalex":"https://openalex.org/W7155403592","doi":"https://doi.org/10.48550/arxiv.2604.19974"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.19974","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19974","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.2604.19974","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5077604549","display_name":"Het Patel","orcid":"https://orcid.org/0000-0001-5321-9178"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Patel, Het","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134386420","display_name":"Tiejin Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Tiejin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134396819","display_name":"Hua Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Hua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125971127","display_name":"Evangelos E. Papalexakis","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Papalexakis, Evangelos E.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134449188","display_name":"Jia Chen","orcid":"https://orcid.org/0009-0009-4788-9576"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jia","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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.6245999932289124,"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":0.6245999932289124,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.08179999887943268,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.05779999867081642,"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/correctness","display_name":"Correctness","score":0.9613999724388123},{"id":"https://openalex.org/keywords/interpretability","display_name":"Interpretability","score":0.789900004863739},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.5627999901771545},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.5340999960899353},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.4602999985218048},{"id":"https://openalex.org/keywords/classifier","display_name":"Classifier (UML)","score":0.42989999055862427},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.4244999885559082}],"concepts":[{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.9613999724388123},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.789900004863739},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6757000088691711},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5633999705314636},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.5627999901771545},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5340999960899353},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.4602999985218048},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.44359999895095825},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.42989999055862427},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4244999885559082},{"id":"https://openalex.org/C33676613","wikidata":"https://www.wikidata.org/wiki/Q13415176","display_name":"Dimension (graph theory)","level":2,"score":0.4242999851703644},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3425999879837036},{"id":"https://openalex.org/C142853389","wikidata":"https://www.wikidata.org/wiki/Q744778","display_name":"Association (psychology)","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2741999924182892},{"id":"https://openalex.org/C2779712033","wikidata":"https://www.wikidata.org/wiki/Q28208841","display_name":"Dissociation rate","level":3,"score":0.26989999413490295},{"id":"https://openalex.org/C111030470","wikidata":"https://www.wikidata.org/wiki/Q1430460","display_name":"Curse of dimensionality","level":2,"score":0.2687000036239624},{"id":"https://openalex.org/C102931765","wikidata":"https://www.wikidata.org/wiki/Q189673","display_name":"Dissociation (chemistry)","level":2,"score":0.2556999921798706},{"id":"https://openalex.org/C83665646","wikidata":"https://www.wikidata.org/wiki/Q42139305","display_name":"Feature vector","level":2,"score":0.25200000405311584}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.19974","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19974","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.2604.19974","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19974","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"can":[3],"be":[4],"uncertain":[5],"yet":[6,10],"correct,":[7],"or":[8,31],"confident":[9],"wrong,":[11],"raising":[12],"the":[13,27,107,146,158],"question":[14],"of":[15,130,160],"whether":[16],"their":[17,21],"output-level":[18],"uncertainty":[19,80,192],"and":[20,47,50,66,104,127,137,148,193,203],"actual":[22],"correctness":[23,46,172,194],"are":[24,82,93,122,154,195],"driven":[25],"by":[26,32],"same":[28],"internal":[29,197],"mechanisms":[30],"distinct":[33,196],"feature":[34,71,152],"populations.":[35],"We":[36],"introduce":[37],"a":[38,133,138,166],"2x2":[39],"framework":[40],"that":[41,73,118,178,191],"partitions":[42],"model":[43,171],"predictions":[44],"along":[45],"confidence":[48],"axes,":[49],"uses":[51],"sparse":[52],"autoencoders":[53],"to":[54,64,124,183],"identify":[55,69],"features":[56,81,92,117,164],"associated":[57],"with":[58,142,199],"each":[59],"dimension":[60],"independently.":[61],"Applying":[62],"this":[63],"Llama-3.1-8B":[65],"Gemma-2-9B,":[67],"we":[68],"three":[70],"populations":[72],"play":[74],"fundamentally":[75],"different":[76],"functional":[77],"roles.":[78],"Pure":[79,90],"functionally":[83,94],"essential:":[84],"suppressing":[85],"them":[86,131],"severely":[87],"degrades":[88],"accuracy.":[89],"incorrectness":[91],"inert:":[95],"despite":[96],"showing":[97],"statistically":[98],"significant":[99],"activation":[100],"differences":[101],"between":[102],"correct":[103],"incorrect":[105],"predictions,":[106],"majority":[108],"produce":[109],"near-zero":[110],"change":[111],"in":[112],"accuracy":[113,135,180],"when":[114],"suppressed.":[115],"Confounded":[116],"encode":[119],"both":[120],"signals":[121],"detrimental":[123],"output":[125],"quality,":[126],"targeted":[128,204],"suppression":[129],"yields":[132],"1.1%":[134],"improvement":[136],"75%":[139],"entropy":[140],"reduction,":[141],"effects":[143],"transferring":[144],"across":[145],"ARC-Challenge":[147],"RACE":[149],"benchmarks.":[150],"The":[151,188],"categories":[153],"also":[155],"informationally":[156],"distinct:":[157],"activations":[159],"just":[161],"3":[162],"confounded":[163],"from":[165,181],"single":[167],"mid-network":[168],"layer":[169],"predict":[170],"(AUROC":[173],"~0.79),":[174],"enabling":[175],"selective":[176],"abstention":[177],"raises":[179],"62%":[182],"81%":[184],"at":[185],"53%":[186],"coverage.":[187],"results":[189],"demonstrate":[190],"phenomena,":[198],"implications":[200],"for":[201],"interpretability":[202],"inference-time":[205],"intervention.":[206]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-24T00:00:00"}