ε-coresets for attention exist of size O(√d e^{ρ+o(ρ)}/ε) for unit-norm keys/values and queries of norm ≤ρ, nearly matching the Ω(√d e^ρ/ε) lower bound.
Wildcat: Near-linear attention in theory and practice
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
abstract
We introduce WildCat, a high-accuracy, low-cost approach to compressing the attention mechanism in neural networks. While attention is a staple of modern network architectures, it is also notoriously expensive to deploy due to resource requirements that scale quadratically with the input sequence length $n$. WildCat avoids these quadratic costs by only attending over a small weighted coreset. Crucially, we select the coreset using a fast but spectrally-accurate subsampling algorithm -- randomly pivoted Cholesky -- and weight the elements optimally to minimise reconstruction error. Remarkably, given bounded inputs, WildCat approximates exact attention with super-polynomial $O(n^{-\sqrt{\log(\log(n))}})$ error decay while running in near-linear $O(n^{1+o(1)})$ time. In contrast, prior practical approximations either lack error guarantees or require quadratic runtime to guarantee such high fidelity. We couple this advance with a GPU-optimized PyTorch implementation and a suite of benchmark experiments demonstrating the benefits of WildCat for image generation, image classification, and language model KV cache compression.
years
2026 2verdicts
UNVERDICTED 2representative citing papers
Express converts non-causal attention approximations to causal versions, achieving log^{3/2}(n)/s error with O(s) memory and O(s^2 log^2(n)) overhead when combined with Thinformer.
citing papers explorer
-
Nearly Optimal Attention Coresets
ε-coresets for attention exist of size O(√d e^{ρ+o(ρ)}/ε) for unit-norm keys/values and queries of norm ≤ρ, nearly matching the Ω(√d e^ρ/ε) lower bound.
-
Express Language Modeling
Express converts non-causal attention approximations to causal versions, achieving log^{3/2}(n)/s error with O(s) memory and O(s^2 log^2(n)) overhead when combined with Thinformer.