Agentic Critical Training

Weize Liu, Minghui Liu, Sy-Tuyen Ho +3

Training large language models (LLMs) as autonomous agents often begins with imitation learning, but it only teaches agents what to do without understanding why: agents never contr…