We adaptes Mamba2's structured mask to 2D scaning and integrates it into the self-attention mechanism of ViTs as an explicit positional encoding.