LGAIDCDec 18, 2025

Kascade: A Practical Sparse Attention Method for Long-Context LLM Inference

arXiv:2512.16391v15 citationsh-index: 47
Originality Incremental advance
AI Analysis

This addresses the bottleneck of attention latency for users of long-context LLMs, offering a practical, incremental improvement with deployment-ready implementation.

The paper tackles the latency problem in long-context LLM inference by proposing Kascade, a training-free sparse attention method that achieves up to 4.1x speedup in decode attention and 2.2x speedup in prefill attention over FlashAttention-3 while closely matching dense attention accuracy on benchmarks like LongBench and AIME-24.

Attention is the dominant source of latency during long-context LLM inference, an increasingly popular workload with reasoning models and RAG. We propose Kascade, a training-free sparse attention method that leverages known observations such as 1) post-softmax attention is intrinsically sparse, and 2) the identity of high-weight keys is stable across nearby layers. Kascade computes exact Top-k indices in a small set of anchor layers, then reuses those indices in intermediate reuse layers. The anchor layers are selected algorithmically, via a dynamic-programming objective that maximizes cross-layer similarity over a development set, allowing easy deployment across models. The method incorporates efficient implementation constraints (e.g. tile-level operations), across both prefill and decode attention. The Top-k selection and reuse in Kascade is head-aware and we show in our experiments that this is critical for high accuracy. Kascade achieves up to 4.1x speedup in decode attention and 2.2x speedup in prefill attention over FlashAttention-3 baseline on H100 GPUs while closely matching dense attention accuracy on long-context benchmarks such as LongBench and AIME-24.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes