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.
Rabitq: Quantizing high-dimensional vectors with a theoretical error bound for approximate nearest neighbor search.Proceedings of the ACM on Management of Data, 2(3):1–27
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
Two randomized Hadamard transforms suffice to make coordinate marginals O(d^{-1/2})-close to Gaussian for most quantization methods, with three needed for vector quantization to match uniform random rotations asymptotically.
Two-block Hadamard rotations match uniform rotations in each coordinate with improving accuracy as dimension grows but differ substantially in their overall distribution, with explicit bounds.
TurboQuant_mse is EDEN with fixed scale S=1 and TurboQuant_prod is inferior to direct unbiased b-bit EDEN due to suboptimal scaling, residual quantization, and biased-unbiased chaining.
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.
-
Quantizing With Randomized Hadamard Transforms: Efficient Heuristic Now Proven
Two randomized Hadamard transforms suffice to make coordinate marginals O(d^{-1/2})-close to Gaussian for most quantization methods, with three needed for vector quantization to match uniform random rotations asymptotically.
-
Approximating Uniform Random Rotations by Two-Block Structured Hadamard Rotations in High Dimensions
Two-block Hadamard rotations match uniform rotations in each coordinate with improving accuracy as dimension grows but differ substantially in their overall distribution, with explicit bounds.
-
A Note on TurboQuant and the Earlier DRIVE/EDEN Line of Work
TurboQuant_mse is EDEN with fixed scale S=1 and TurboQuant_prod is inferior to direct unbiased b-bit EDEN due to suboptimal scaling, residual quantization, and biased-unbiased chaining.