1 paper
Liting Lin, Boxi Yu, Yuzhong Zhang +3
Conversational LLM agents can cause real-world harm when their internal workflows fail, such as completing a transaction without confirmation. Testing these state-dependent failure…