While directly applying vision-RWKV yields suboptimal results due to insufficient local feature exploration and disrupted spatial continuity, we propose a novel nested structure, Zigzag RWKV-in-RWKV ...