arXiv stat.ML
· Papers
Native Multi-Dimensional Subquadratic Operators via Input Dependent Long Convolutions
arXiv:2607.19378v1 Announce Type: cross Abstract: Subquadratic alternatives to attention require compromises when applied to multi-dimensional data: standard convolutions lack global receptive fields and input dependency, while recurrent models require rasterizing data such as images, volumes, and partial differential