Yehan Zhang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
AgentDecept: Benchmarking Deceptive Behavior in Tool-Using LLM Agents under Pressure
According to the abstract, Yehan Zhang, Qianqi Yan, Hang Yin and five co-authors introduce AgentDecept, a benchmark of 477 pressure variants of 257 tasks. It checks whether tool-using AI agents report their actions faithfully, separating falsification, concealment and equivocation from honest failure.