Paperlane

no. 15 · Monday 16 November 2026

Express Language Modeling

Albert Gong, Annabelle Michael Carrell, Raaz Dwivedi and Lester Mackey

math.ST · v5

Abstract

We introduce a new tool, Express, for converting a non-causal attention approximation into a causal approximation with matching approximation guarantees. When combined with the state-of-the-art Thinformer approximation, Express improves upon the best known causal attention guarantees, delivering $\log^{3/2}(n)/s$ approximation error with only $O(s)$ memory and $O(s^2 \log^2(n))$ compression overhead for a sequence of length $n$. We pair these developments with an efficient I/O-aware Triton implementation, demonstrate substantial speedups over FlashAttention 2, and use Express to overcome four resource bottlenecks in the language modeling pipeline: long-context prefill, KV cache compression, long-form memory-constrained decoding, and long-form compute-constrained decoding.

fields
Statistics Theory
pdf ↗ page arXiv:2606.10944 ↗