FraudBench: Stress-Testing Policy-Grounded Banking Agents Against Adaptive Fraud
FraudBench is introduced, an executable benchmark built on the $\tau^2$-bench dual-control framework and the $\tau$-Knowledge banking environment that tests whether a policy-grounded banking agent safely acts when a caller manipulates identity, authorization, and trust over a conversation.
Dheeraj Mohandas Pai, Lu Xian
· 0 citations