Dimension-Free Minimax Rates for Learning Pairwise Interactions in Attention-Style Models
arXiv:2510.11789v1 Announce Type: new
Abstract: We study the convergence rate of learning pairwise interactions in single-layer attention-style models, where tokens interact through a weight matrix and a non-linear activation function. We prove that the minimax rate is $M^{-frac{2beta}{2beta+1}}$ with $M$ being the sample size, depending only on the smoothness $beta$ of the activation, and crucially independent of token count, ambient dimension, or rank of the weight matrix. These results highlight a fundamental dimension-free statistical efficiency of attention-style nonlocal models, even when the weight matrix and activation are not separately identifiable and provide a theoretical understanding of the attention mechanism and its training.
Abstract: We study the convergence rate of learning pairwise interactions in single-layer attention-style models, where tokens interact through a weight matrix and a non-linear activation function. We prove that the minimax rate is $M^{-frac{2beta}{2beta+1}}$ with $M$ being the sample size, depending only on the smoothness $beta$ of the activation, and crucially independent of token count, ambient dimension, or rank of the weight matrix. These results highlight a fundamental dimension-free statistical efficiency of attention-style nonlocal models, even when the weight matrix and activation are not separately identifiable and provide a theoretical understanding of the attention mechanism and its training.
Shai Zucker, Xiong Wang, Fei Lu, Inbar Seroussi
Go to original source