An Illustrated Summary of "Robust Agents Learn Causal World Model" — AI Alignment Forum