{"id":"https://openalex.org/W7167202790","doi":"https://doi.org/10.48550/arxiv.2607.01793","title":"Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification","display_name":"Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification","publication_year":2026,"publication_date":"2026-07-02","ids":{"openalex":"https://openalex.org/W7167202790","doi":"https://doi.org/10.48550/arxiv.2607.01793"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.01793","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.01793","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2607.01793","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103258005","display_name":"Yunhao Feng","orcid":"https://orcid.org/0009-0005-3042-5266"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Yunhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136208728","display_name":"Ruixiao Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Ruixiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139999841","display_name":"Ming Wen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wen, Ming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139979596","display_name":"Qinqin He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Qinqin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139963907","display_name":"Yanming Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Yanming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049726719","display_name":"Yan Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140009202","display_name":"Yutao Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yutao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101864441","display_name":"Jialuo Chen","orcid":"https://orcid.org/0000-0003-4322-4285"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jialuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Xu, Zhuoer","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Zhuoer","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067405541","display_name":"Xiaohu Du","orcid":"https://orcid.org/0000-0003-4455-3128"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Xiaohu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101852566","display_name":"Jianan Ma","orcid":"https://orcid.org/0009-0007-0448-1218"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Jianan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133722462","display_name":"Zixing Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Zixing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140007287","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/A5140014893","display_name":"Yunhao Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Yunhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140007221","display_name":"Xinhao Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Xinhao","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/T13295","display_name":"Safety Systems Engineering in Autonomy","score":0.39570000767707825,"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"}},"topics":[{"id":"https://openalex.org/T13295","display_name":"Safety Systems Engineering in Autonomy","score":0.39570000767707825,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.16699999570846558,"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.08820000290870667,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/executable","display_name":"Executable","score":0.8575999736785889},{"id":"https://openalex.org/keywords/software-testing","display_name":"Software testing","score":0.44190001487731934},{"id":"https://openalex.org/keywords/predicate","display_name":"Predicate (mathematical logic)","score":0.41600000858306885},{"id":"https://openalex.org/keywords/symbolic-execution","display_name":"Symbolic execution","score":0.32710000872612},{"id":"https://openalex.org/keywords/software","display_name":"Software","score":0.31310001015663147},{"id":"https://openalex.org/keywords/formal-verification","display_name":"Formal verification","score":0.30489999055862427},{"id":"https://openalex.org/keywords/model-checking","display_name":"Model checking","score":0.30160000920295715}],"concepts":[{"id":"https://openalex.org/C160145156","wikidata":"https://www.wikidata.org/wiki/Q778586","display_name":"Executable","level":2,"score":0.8575999736785889},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7071999907493591},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.48350000381469727},{"id":"https://openalex.org/C2984328558","wikidata":"https://www.wikidata.org/wiki/Q188522","display_name":"Software testing","level":3,"score":0.44190001487731934},{"id":"https://openalex.org/C140146324","wikidata":"https://www.wikidata.org/wiki/Q1144319","display_name":"Predicate (mathematical logic)","level":2,"score":0.41600000858306885},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.3418999910354614},{"id":"https://openalex.org/C2779639559","wikidata":"https://www.wikidata.org/wiki/Q7661178","display_name":"Symbolic execution","level":3,"score":0.32710000872612},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.31310001015663147},{"id":"https://openalex.org/C111498074","wikidata":"https://www.wikidata.org/wiki/Q173326","display_name":"Formal verification","level":2,"score":0.30489999055862427},{"id":"https://openalex.org/C110251889","wikidata":"https://www.wikidata.org/wiki/Q1569697","display_name":"Model checking","level":2,"score":0.30160000920295715},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.2906999886035919},{"id":"https://openalex.org/C2779907942","wikidata":"https://www.wikidata.org/wiki/Q7239630","display_name":"Predicate abstraction","level":3,"score":0.28999999165534973},{"id":"https://openalex.org/C116253237","wikidata":"https://www.wikidata.org/wiki/Q1437424","display_name":"Formal specification","level":2,"score":0.28529998660087585},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.28200000524520874},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.2761000096797943},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.27309998869895935},{"id":"https://openalex.org/C53942775","wikidata":"https://www.wikidata.org/wiki/Q1211721","display_name":"Code coverage","level":3,"score":0.25850000977516174},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.25699999928474426},{"id":"https://openalex.org/C132835097","wikidata":"https://www.wikidata.org/wiki/Q7663745","display_name":"System safety","level":2,"score":0.2556000053882599},{"id":"https://openalex.org/C97686452","wikidata":"https://www.wikidata.org/wiki/Q7604153","display_name":"Static analysis","level":2,"score":0.2549999952316284}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.01793","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.01793","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2607.01793","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.01793","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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":[{"score":0.6231943368911743,"id":"https://metadata.un.org/sdg/9","display_name":"Industry, innovation and infrastructure"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"LLM":[0],"agents":[1,39,61,123],"increasingly":[2],"perform":[3],"autonomous":[4],"actions":[5],"through":[6,62],"external":[7],"tools,":[8],"leading":[9],"to":[10,36],"complex":[11],"and":[12,24,73,84,109,146,209],"evolving":[13,215],"safety":[14,18,22,50,80,96,102,168,187,211],"risks.":[15],"However,":[16],"existing":[17],"testing":[19,51,57,203],"targets":[20],"expert-designed":[21],"violations,":[23],"the":[25],"corresponding":[26],"outcomes":[27,142],"are":[28],"evaluated":[29],"by":[30],"hard-coded":[31],"rules,":[32],"making":[33],"them":[34],"costly":[35],"extend":[37],"as":[38],"evolve.":[40],"To":[41],"this":[42],"end,":[43],"we":[44,180],"present":[45],"Vera,":[46],"an":[47],"end-to-end":[48],"automated":[49],"framework":[52],"that":[53,200],"instantiates":[54],"software":[55],"engineering":[56],"principles":[58],"for":[59,207],"non-deterministic":[60],"a":[63,68,100,104,110,128],"three-stage,":[64],"self-reinforcing":[65],"pipeline.":[66],"First,":[67],"literature-driven":[69],"exploration":[70],"continuously":[71],"discovers":[72],"structures":[74],"emerging":[75],"risks":[76],"into":[77],"taxonomies":[78],"of":[79,213],"risks,":[81],"attack":[82,172],"methods,":[83],"tool":[85],"execution":[86,120,195],"environments.":[87],"Second,":[88],"combinatorial":[89],"composition":[90],"across":[91,193],"taxonomy":[92],"dimensions":[93],"produces":[94],"executable":[95,186,202],"cases,":[97],"each":[98],"specifying":[99],"concrete":[101],"goal,":[103],"programmatically":[105],"constructed":[106],"initial":[107],"state,":[108],"deterministic":[111],"verification":[112],"predicate":[113],"grounded":[114],"in":[115,124],"observable":[116],"artifacts.":[117],"Third,":[118],"adaptive":[119],"runs":[121],"heterogeneous":[122],"isolated":[125],"sandboxes":[126],"where":[127],"control":[129],"agent":[130,159],"steers":[131],"multi-turn":[132],"interaction":[133],"based":[134],"on":[135,156],"runtime":[136],"observations,":[137],"while":[138],"evidence-grounded":[139],"verifiers":[140],"judge":[141],"from":[143],"environment":[144],"state":[145],"tool-call":[147],"evidence":[148],"rather":[149],"than":[150],"model":[151],"self-report.":[152],"We":[153],"evaluate":[154],"Vera":[155],"four":[157],"production":[158],"frameworks":[160],"(OpenClaw,":[161],"Hermes,":[162],"Codex,":[163],"Claude":[164],"Code),":[165],"revealing":[166],"substantial":[167],"weaknesses,":[169],"with":[170],"average":[171],"success":[173],"rates":[174],"reaching":[175],"93.9\\%":[176],"under":[177],"multi-channel":[178],"attacks;":[179],"also":[181],"release":[182],"Vera-Bench,":[183],"comprising":[184],"1600":[185],"cases":[188],"spanning":[189],"124":[190],"risk":[191],"categories":[192],"three":[194],"settings.":[196],"These":[197],"results":[198],"indicate":[199],"modular,":[201],"infrastructure":[204],"is":[205,222],"essential":[206],"rigorous":[208],"maintainable":[210],"evaluation":[212],"rapidly":[214],"agentic":[216],"systems":[217],"at":[218,225],"scale.":[219],"The":[220],"code":[221],"publicly":[223],"available":[224],"https://github.com/Yunhao-Feng/Vera.":[226]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-04T00:00:00"}
