{"id":"https://openalex.org/W7163139661","doi":"https://doi.org/10.48550/arxiv.2606.00819","title":"Mitigating Hallucinations in Large Language Models Via Decoder Layer Skipping","display_name":"Mitigating Hallucinations in Large Language Models Via Decoder Layer Skipping","publication_year":2026,"publication_date":"2026-05-30","ids":{"openalex":"https://openalex.org/W7163139661","doi":"https://doi.org/10.48550/arxiv.2606.00819"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.00819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.00819","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.00819","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5058466211","display_name":"Hanze Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Hanze","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082693976","display_name":"Jinhao You","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"You, Jinhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137655859","display_name":"Yichen Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Yichen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137664927","display_name":"Kai Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Kai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071296285","display_name":"Shuangyang Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Shuangyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5128654904","display_name":"Xiande Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Xiande","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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.19120000302791595,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.19120000302791595,"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/T10028","display_name":"Topic Modeling","score":0.10339999943971634,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.04699999839067459,"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/decoding-methods","display_name":"Decoding methods","score":0.7648000121116638},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.6262999773025513},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5857999920845032},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.43230000138282776},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.3864000141620636},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.3817000091075897},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.3540000021457672}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7871999740600586},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.7648000121116638},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6262999773025513},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5857999920845032},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.43230000138282776},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4268999993801117},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3864000141620636},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.3817000091075897},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.3540000021457672},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.34599998593330383},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2969000041484833},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.29339998960494995},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.27649998664855957},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.26820001006126404},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26660001277923584},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.265500009059906},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25859999656677246},{"id":"https://openalex.org/C2779696439","wikidata":"https://www.wikidata.org/wiki/Q7512811","display_name":"Signature (topology)","level":2,"score":0.25690001249313354},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.25619998574256897},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.25440001487731934}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.00819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.00819","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.00819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.00819","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":"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":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"achieved":[5],"strong":[6],"performance":[7],"across":[8,145],"diverse":[9,146],"natural":[10],"language":[11,173],"tasks,":[12],"yet":[13],"their":[14,130],"outputs":[15],"often":[16],"suffer":[17],"from":[18,48,109],"hallucinations":[19,44,155],"--":[20],"content":[21],"that":[22,43,66,79,151],"is":[23,87],"misaligned":[24],"with":[25,133],"factual":[26],"information.":[27],"In":[28],"this":[29,54],"work,":[30],"we":[31,56],"conduct":[32],"a":[33,62,98,161],"comprehensive":[34],"layer-wise":[35],"analysis":[36],"of":[37,83,93,171],"the":[38,76,80,103,117,169],"decoding":[39,64,165],"process":[40],"and":[41,148,156,163],"reveal":[42],"tend":[45],"to":[46,71,90],"originate":[47],"deeper":[49],"decoder":[50,111],"layers.":[51],"To":[52],"address":[53],"issue,":[55],"introduce":[57],"\\textbf{DeLask}":[58],"(\\textbf{De}coder":[59],"\\textbf{La}yer":[60],"\\textbf{Sk}ipping),":[61],"novel":[63],"framework":[65,166],"dynamically":[67],"skips":[68],"layers":[69,115,125],"prone":[70],"producing":[72],"hallucinations.":[73],"DeLask":[74,127,152],"leverages":[75],"theoretical":[77],"insight":[78],"forward":[81],"computation":[82],"an":[84],"$L$-layer":[85],"Transformer":[86],"conditionally":[88],"equivalent":[89],"$L$":[91],"steps":[92],"gradient":[94],"descent.":[95],"We":[96],"define":[97],"\\emph{driftance":[99],"value}":[100],"by":[101],"computing":[102],"cosine":[104],"similarity":[105],"between":[106],"gradients":[107],"derived":[108],"consecutive":[110],"steps,":[112],"identifying":[113],"problematic":[114],"when":[116],"descent":[118],"direction":[119],"reverses.":[120],"Rather":[121],"than":[122],"discarding":[123],"such":[124],"entirely,":[126],"partially":[128],"aggregates":[129],"hidden":[131],"states":[132],"preceding":[134],"layers,":[135],"thereby":[136],"preserving":[137],"consistency":[138],"while":[139],"suppressing":[140],"erroneous":[141],"signals.":[142],"Extensive":[143],"experiments":[144],"LLMs":[147],"benchmarks":[149],"demonstrate":[150],"consistently":[153],"mitigates":[154],"enhances":[157],"overall":[158],"reliability,":[159],"providing":[160],"lightweight":[162],"generalizable":[164],"for":[167],"improving":[168],"robustness":[170],"large-scale":[172],"models.":[174]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-03T00:00:00"}