{"id":"https://openalex.org/W7133192295","doi":"https://doi.org/10.48550/arxiv.2602.24027","title":"GuardAlign: Test-time Safety Alignment in Multimodal Large Language Models","display_name":"GuardAlign: Test-time Safety Alignment in Multimodal Large Language Models","publication_year":2026,"publication_date":"2026-02-27","ids":{"openalex":"https://openalex.org/W7133192295","doi":"https://doi.org/10.48550/arxiv.2602.24027"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.24027","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":"https://openalex.org/A5127808960","display_name":"Xingyu Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Xingyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009738035","display_name":"Beier Zhu","orcid":"https://orcid.org/0000-0002-7900-6979"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Beier","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127792508","display_name":"Junfeng Fang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Junfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127814106","display_name":"Shuo Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Shuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127844390","display_name":"Yin Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100389031","display_name":"Xiang Wang","orcid":"https://orcid.org/0000-0002-5211-7206"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Xiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127836546","display_name":"Xiangnan He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Xiangnan","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":0.4388999938964844,"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":0.4388999938964844,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.2897000014781952,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0640999972820282,"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/identification","display_name":"Identification (biology)","score":0.5349000096321106},{"id":"https://openalex.org/keywords/prefix","display_name":"Prefix","score":0.47609999775886536},{"id":"https://openalex.org/keywords/measure","display_name":"Measure (data warehouse)","score":0.4397999942302704},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.3601999878883362},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.35659998655319214},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.28209999203681946}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7405999898910522},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.5349000096321106},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.47609999775886536},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.4397999942302704},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4059000015258789},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3601999878883362},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.35659998655319214},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.33070001006126404},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.28209999203681946},{"id":"https://openalex.org/C132835097","wikidata":"https://www.wikidata.org/wiki/Q7663745","display_name":"System safety","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2806999981403351},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2680000066757202},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.25369998812675476},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.2524999976158142}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.24027","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.2602.24027","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.24027","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":"pmh:doi:10.48550/arxiv.2602.24027","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"vision-language":[1,9],"models":[2],"(LVLMs)":[3],"have":[4],"achieved":[5],"remarkable":[6],"progress":[7],"in":[8,41],"reasoning":[10],"tasks,":[11],"yet":[12],"ensuring":[13,108],"their":[14],"safety":[15,30,46,67,100,110],"remains":[16],"a":[17,57],"critical":[18],"challenge.":[19],"Recent":[20],"input-side":[21],"defenses":[22],"detect":[23],"unsafe":[24,80,127],"images":[25],"with":[26],"CLIP":[27],"and":[28,44,79],"prepend":[29],"prefixes":[31,101],"to":[32,72,132,146],"prompts,":[33],"but":[34],"they":[35],"still":[36],"suffer":[37],"from":[38,144],"inaccurate":[39],"detection":[40,68],"complex":[42],"scenes":[43],"unstable":[45],"signals":[47,111],"during":[48],"decoding.":[49],"To":[50],"address":[51],"these":[52],"issues,":[53],"we":[54],"propose":[55],"GuardAlign,":[56],"training-free":[58],"defense":[59],"framework":[60],"that":[61,109,124],"integrates":[62],"two":[63],"strategies.":[64],"First,":[65],"OT-enhanced":[66],"leverages":[69],"optimal":[70],"transport":[71],"measure":[73],"distribution":[74],"distances":[75],"between":[76],"image":[77],"patches":[78],"semantics,":[81],"enabling":[82],"accurate":[83],"identification":[84],"of":[85,99],"malicious":[86],"regions":[87],"without":[88],"additional":[89],"computational":[90],"cost.":[91],"Second,":[92],"cross-modal":[93],"attentive":[94],"calibration":[95],"strengthens":[96],"the":[97],"influence":[98],"by":[102,130],"adaptively":[103],"reallocating":[104],"attention":[105],"across":[106],"layers,":[107],"remain":[112],"consistently":[113],"activated":[114],"throughout":[115],"generation.":[116],"Extensive":[117],"evaluations":[118],"on":[119,134,142],"six":[120],"representative":[121],"MLLMs":[122],"demonstrate":[123],"GuardAlign":[125],"reduces":[126],"response":[128],"rates":[129],"up":[131],"39%":[133],"SPA-VL,":[135],"while":[136],"preserving":[137],"utility,":[138],"achieving":[139],"an":[140],"improvement":[141],"VQAv2":[143],"78.51%":[145],"79.21%.":[147]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-03T00:00:00"}