1 paper
KuanChao Chu, Yi-Pei Chen, Hideki Nakayama
Controlling diversity in LLM-agent simulations is essential for balancing stability in structured tasks with variability in open-ended interactions. However, we observe that dialog…