We released our paper, RoPeSLR: 3D RoPE-driven Sparse-LowRank Attention for Efficient Diffusion Transformers.