List your company

Policy Violation Detection on SafeAgentBench

234True PositivesVIGIL
221.52224.76228231.24Jun 25, 2026

Evaluation results

MethodLinks
VIGILTests=5462026.06234498.394.496.3
AgentSpecTests=5462026.06231299.193.196
LLM-as-judgeTests=5462026.062224184.489.586.9