AsyncTLS: Efficient Generative LLM Inference with Asynchronous Two-level Sparse Attention

TL;DR

AsyncTLS introduces a hierarchical sparse attention mechanism with asynchronous offloading to enable efficient long-context inference in large language models, balancing accuracy and computational efficiency.

Contribution

It combines coarse-grained block filtering with fine-grained token selection and an asynchronous offloading engine, achieving high accuracy and significant speedups.

Findings

Achieves accuracy comparable to full attention.

Delivers 1.2x - 10.0x operator speedups.

Provides 1.3x - 4.7x end-to-end throughput improvements.

Abstract

Long-context inference in LLMs faces the dual challenges of quadratic attention complexity and prohibitive KV cache memory. While token-level sparse attention offers superior accuracy, its indexing overhead is costly; block-level methods improve efficiency but sacrifice precision. We propose AsyncTLS, a hierarchical sparse attention system that combines coarse-grained block filtering with fine-grained token selection to balance accuracy and efficiency, coupled with an asynchronous offloading engine that overlaps KV cache transfers with computation via temporal locality exploitation. Evaluated on Qwen3 and GLM-4.7-Flash across GQA, and MLA architectures, AsyncTLS achieves accuracy comparable to full attention while delivering 1.2x - 10.0x operator speedups and 1.3x - 4.7x end-to-end throughput improvements on 48k - 96k contexts.