Hao Liang

We propose a generalizable multi-agent reinforcement learning framework exploiting causal structures, enabling efficient adaptation across networked systems under domain shifts, validated by theoretical guarantees and numerical experiments