Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
2
Recurrent token mixing via GTR achieves 58.9 AP on COCO with sub-2ms latency, offering a practical, efficient alternative to quadratic self-attention for high-resolution dense prediction.
Laplacian kernels offer a theoretically sound and empirically superior alternative to Gaussian kernels for linear attention, finally bridging the gap between efficient computation and expressive power in Transformers.
