{"id":"https://openalex.org/W7160134058","doi":"https://doi.org/10.48550/arxiv.2605.00689","title":"ML-Bench&Guard: Policy-Grounded Multilingual Safety Benchmark and Guardrail for Large Language Models","display_name":"ML-Bench&Guard: Policy-Grounded Multilingual Safety Benchmark and Guardrail for Large Language Models","publication_year":2026,"publication_date":"2026-05-01","ids":{"openalex":"https://openalex.org/W7160134058","doi":"https://doi.org/10.48550/arxiv.2605.00689"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.00689","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.00689","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.2605.00689","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135224247","display_name":"Yunhan Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Yunhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135124426","display_name":"Zhaorun Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Zhaorun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135194576","display_name":"Xingjun Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Xingjun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135143624","display_name":"Yu-Gang Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Yu-Gang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135171910","display_name":"Bo Li","orcid":"https://orcid.org/0000-0002-5131-8639"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Bo","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":0.5041000247001648,"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":0.5041000247001648,"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/T13295","display_name":"Safety Systems Engineering in Autonomy","score":0.09640000015497208,"subfield":{"id":"https://openalex.org/subfields/2213","display_name":"Safety, Risk, Reliability and Quality"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.05869999900460243,"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/benchmark","display_name":"Benchmark (surveying)","score":0.6186000108718872},{"id":"https://openalex.org/keywords/bridge","display_name":"Bridge (graph theory)","score":0.5853999853134155},{"id":"https://openalex.org/keywords/multilingualism","display_name":"Multilingualism","score":0.36070001125335693},{"id":"https://openalex.org/keywords/cultural-diversity","display_name":"Cultural diversity","score":0.3449999988079071},{"id":"https://openalex.org/keywords/compliance","display_name":"Compliance (psychology)","score":0.32919999957084656},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.2856000065803528}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7002999782562256},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.6186000108718872},{"id":"https://openalex.org/C100776233","wikidata":"https://www.wikidata.org/wiki/Q2532492","display_name":"Bridge (graph theory)","level":2,"score":0.5853999853134155},{"id":"https://openalex.org/C2780035574","wikidata":"https://www.wikidata.org/wiki/Q30081","display_name":"Multilingualism","level":2,"score":0.36070001125335693},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3582000136375427},{"id":"https://openalex.org/C125209646","wikidata":"https://www.wikidata.org/wiki/Q1338878","display_name":"Cultural diversity","level":2,"score":0.3449999988079071},{"id":"https://openalex.org/C2781460075","wikidata":"https://www.wikidata.org/wiki/Q1399332","display_name":"Compliance (psychology)","level":2,"score":0.32919999957084656},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.29319998621940613},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2856000065803528},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.2759000062942505},{"id":"https://openalex.org/C2986201853","wikidata":"https://www.wikidata.org/wiki/Q11042","display_name":"Cultural issues","level":3,"score":0.2709999978542328},{"id":"https://openalex.org/C47602998","wikidata":"https://www.wikidata.org/wiki/Q2588869","display_name":"Language barrier","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.263700008392334},{"id":"https://openalex.org/C2778883600","wikidata":"https://www.wikidata.org/wiki/Q2390977","display_name":"Language proficiency","level":2,"score":0.25459998846054077}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.00689","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.00689","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.2605.00689","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.00689","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":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.7073123455047607}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"As":[0],"Large":[1,117],"Language":[2,118],"Models":[3],"(LLMs)":[4],"are":[5,90],"increasingly":[6],"deployed":[7],"in":[8,13],"cross-linguistic":[9],"contexts,":[10],"ensuring":[11],"safety":[12,67,99,126,170],"diverse":[14],"regulatory":[15],"and":[16,34,45,54,82,103,128,144,172,175,187,196],"cultural":[17,55],"environments":[18],"has":[19,133],"become":[20],"a":[21,64,115,145],"critical":[22],"challenge.":[23],"However,":[24],"existing":[25,168],"multilingual":[26,66,98,125,169,199],"benchmarks":[27,171],"largely":[28],"rely":[29],"on":[30,110],"general":[31],"risk":[32,80],"taxonomies":[33],"machine":[35],"translation,":[36],"which":[37],"confines":[38],"guardrail":[39,121,164,200],"models":[40],"to":[41,49,93],"these":[42,59],"predefined":[43],"categories":[44,81],"hinders":[46],"their":[47],"ability":[48],"align":[50],"with":[51,154],"region-specific":[52],"regulations":[53],"nuances.":[56],"To":[57],"bridge":[58],"gaps,":[60],"we":[61,112],"introduce":[62],"ML-Bench,":[63,111,174],"policy-grounded":[65],"benchmark":[68],"covering":[69],"14":[70],"languages.":[71,108],"ML-Bench":[72,186],"is":[73],"constructed":[74],"directly":[75,91],"from":[76,86],"regional":[77],"regulations,":[78],"where":[79],"fine-grained":[83],"rules":[84],"derived":[85],"jurisdiction-specific":[87],"legal":[88],"texts":[89],"used":[92],"guide":[94],"the":[95,192],"generation":[96],"of":[97,194],"data,":[100],"enabling":[101],"culturally":[102,197],"legally":[104],"aligned":[105,198],"evaluation":[106],"across":[107,166],"Building":[109],"develop":[113],"ML-Guard,":[114],"Diffusion":[116],"Model":[119],"(dLLM)-based":[120],"model":[122,139,149],"that":[123,177,185],"supports":[124],"judgment":[127],"policy-conditioned":[129],"compliance":[130,152],"assessment.":[131],"ML-Guard":[132,178,188],"two":[134],"variants,":[135],"one":[136],"1.5B":[137],"lightweight":[138],"for":[140,150],"fast":[141],"`safe/unsafe'":[142],"checking":[143,153],"more":[146],"capable":[147],"7B":[148],"customized":[151],"detailed":[155],"explanations.":[156],"We":[157,183],"conduct":[158],"extensive":[159],"experiments":[160],"against":[161],"11":[162],"strong":[163],"baselines":[165],"6":[167],"our":[173],"show":[176],"consistently":[179],"outperforms":[180],"prior":[181],"methods.":[182],"hope":[184],"can":[189],"help":[190],"advance":[191],"development":[193],"regulation-aware":[195],"systems.":[201]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-05T00:00:00"}