1 paper
Chaoran Chen, Vy Nguyen, Ziji Zhang +7
Tool-using LLM agents are typically developed and post-trained in controlled environments with reliable tools, but deployment exposes them to tool failures, API and schema drift, a…