1 paper
Nunzio Lorè, Hongan Zhu, Babak Heydari
Large language models are increasingly deployed as interacting agents, making the persistence of their action policies across repeated interaction critical for reliable multi-agent…