Testing and Understanding Erroneous Planning in LLM Agents through Synthesized User Inputs

Open in new window