{"id":"https://openalex.org/W7127402842","doi":"https://doi.org/10.48550/arxiv.2602.01167","title":"Do All Individual Layers Help? An Empirical Study of Task-Interfering Layers in Vision-Language Models","display_name":"Do All Individual Layers Help? An Empirical Study of Task-Interfering Layers in Vision-Language Models","publication_year":2026,"publication_date":"2026-02-01","ids":{"openalex":"https://openalex.org/W7127402842","doi":"https://doi.org/10.48550/arxiv.2602.01167"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.01167","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","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":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Liu, Zhiming","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Liu, Zhiming","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124936318","display_name":"Yujie Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Yujie","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124899893","display_name":"Lei Feng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Lei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102814914","display_name":"Xiu Su","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Su, Xiu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114803721","display_name":"Xiaobo Xia","orcid":"https://orcid.org/0000-0003-3615-0919"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xia, Xiaobo","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5118451722","display_name":"Weili Guan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guan, Weili","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124842349","display_name":"Zeke Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Zeke","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5124930361","display_name":"Shuo Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Shuo","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":8,"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":0.8264999985694885,"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":0.8264999985694885,"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":0.050200000405311584,"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":0.01720000058412552,"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/modularity","display_name":"Modularity (biology)","score":0.6111000180244446},{"id":"https://openalex.org/keywords/layer","display_name":"Layer (electronics)","score":0.5845000147819519},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5665000081062317},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.501800000667572},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.49149999022483826},{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.46309998631477356},{"id":"https://openalex.org/keywords/mechanism","display_name":"Mechanism (biology)","score":0.4196999967098236},{"id":"https://openalex.org/keywords/downstream","display_name":"Downstream (manufacturing)","score":0.41519999504089355}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7174000144004822},{"id":"https://openalex.org/C2779478453","wikidata":"https://www.wikidata.org/wiki/Q6889748","display_name":"Modularity (biology)","level":2,"score":0.6111000180244446},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.5845000147819519},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5665000081062317},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.501800000667572},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.49149999022483826},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.46309998631477356},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.4196999967098236},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.41519999504089355},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.4104999899864197},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4011000096797943},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3885999917984009},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.36800000071525574},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.367900013923645},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.34150001406669617},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3400999903678894},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.31189998984336853},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30390000343322754},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.2946000099182129},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.290800005197525},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.27070000767707825},{"id":"https://openalex.org/C135798126","wikidata":"https://www.wikidata.org/wiki/Q2167279","display_name":"Top-down and bottom-up design","level":2,"score":0.2596000134944916},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.25459998846054077}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.01167","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.01167","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.01167","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":"pmh:doi:10.48550/arxiv.2602.01167","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","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":{"Current":[0],"VLMs":[1,228],"have":[2],"demonstrated":[3],"capabilities":[4,146,238],"across":[5,101,199],"a":[6,14,34,131,134,175,190,231],"wide":[7],"range":[8],"of":[9,108,125,130,224],"multimodal":[10],"tasks.":[11,28,59],"Typically,":[12],"in":[13,77,160,212,226],"pretrained":[15,227],"VLM,":[16],"all":[17],"layers":[18,53,65,138],"are":[19],"engaged":[20],"by":[21,39,156,166,214],"default":[22],"to":[23,80,216,235],"make":[24],"predictions":[25],"on":[26,33,47,86,127,208],"downstream":[27,58,113],"We":[29,60,116],"find":[30],"that":[31,51,111,180],"intervening":[32,85,126],"single":[35],"layer,":[36],"such":[37],"as":[38,154],"zeroing":[40],"its":[41],"parameters,":[42],"can":[43,98],"improve":[44],"the":[45,75,81,106,123,157,185,209],"performance":[46,78,198],"certain":[48],"tasks,":[49],"indicating":[50,105],"some":[52],"hinder":[54],"rather":[55],"than":[56],"help":[57],"systematically":[61],"investigate":[62],"how":[63],"individual":[64],"influence":[66],"different":[67],"tasks":[68,143],"via":[69],"layer":[70,88,129,152,188],"intervention.":[71],"Specifically,":[72],"we":[73,169],"measure":[74],"change":[76],"relative":[79],"base":[82],"model":[83],"after":[84],"each":[87,128],"and":[89,103,183,202,229],"observe":[90],"improvements":[91],"when":[92],"bypassing":[93],"specific":[94],"layers.":[95],"This":[96],"improvement":[97],"be":[99,246],"generalizable":[100],"models":[102,201],"datasets,":[104,203],"presence":[107],"Task-Interfering":[109],"Layers":[110],"harm":[112],"tasks'":[114],"performance.":[115],"introduce":[117],"Task-Layer":[118],"Interaction":[119],"Vector,":[120],"which":[121],"quantifies":[122],"effect":[124],"VLM":[132],"given":[133,191],"task.":[135,192],"These":[136],"task-interfering":[137],"exhibit":[139],"task-specific":[140],"sensitivity":[141],"patterns:":[142],"requiring":[144],"similar":[145],"show":[147],"consistent":[148],"response":[149],"trends":[150],"under":[151],"interventions,":[153],"evidenced":[155],"high":[158],"similarity":[159],"their":[161],"task-layer":[162],"interaction":[163],"vectors.":[164],"Inspired":[165],"these":[167],"findings,":[168],"propose":[170],"TaLo":[171,196],"(Task-Adaptive":[172],"Layer":[173],"Knockout),":[174],"training-free,":[176],"test-time":[177],"adaptation":[178],"method":[179],"dynamically":[181],"identifies":[182],"bypasses":[184],"most":[186],"interfering":[187],"for":[189],"Without":[193],"parameter":[194],"updates,":[195],"improves":[197],"various":[200],"including":[204],"boosting":[205],"Qwen-VL's":[206],"accuracy":[207],"Maps":[210],"task":[211],"ScienceQA":[213],"up":[215],"16.6%.":[217],"Our":[218],"work":[219],"reveals":[220],"an":[221],"unexpected":[222],"form":[223],"modularity":[225],"provides":[230],"plug-and-play,":[232],"training-free":[233],"mechanism":[234],"unlock":[236],"hidden":[237],"at":[239],"inference":[240],"time.":[241],"The":[242],"source":[243],"code":[244],"will":[245],"publicly":[247],"available.":[248]},"counts_by_year":[],"updated_date":"2026-04-04T16:13:02.066488","created_date":"2026-02-04T00:00:00"}