1 paper
Kailong Ren, Fubo Sun, Jiachen Liu +18
As large language model (LLM) agents are applied to longer tasks, they increasingly modify workspace state across multiple rounds of iteration. However, agents typically observe on…