multi agent simulations

Towards Human-Agent Communication via the Information Bottleneck Principle