AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

FSA: An Alternative Efficient Implementation of Native Sparse Attention Kernel

Ran Yan, Youhe Jiang, Zhuoming Chen, Haohui Mai, Beidi Chen, Binhang Yuan

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
invert kernel loop order (loop KV blocks outer, queries inner)NSA kernel · small GQA groups (g<8)kernel latency: up to 3.5x, avg 1.6x lower latency (abstract)Improved
use FSA instead of NSA kernelend-to-end training · Llama3-8B, Qwen3-14B, Qwen2.5-32Btraining latency: avg 1.09x, up to 1.25xImproved
use FSA instead of NSA kernelinference prefill · LLMs, 32K-64Kprefill latency: avg 1.11x, up to 1.36xImproved
use FSA instead of NSA kerneldecoding · LLM inferencedecoding latencyNo clear effect
use FSA at GQA group size 8kernel · H20 prefill 128K, GQA=8latency: 0.45 vs 0.45 sNo clear effect
disable inner-loop optimization (ablation)FSA kernel · kernel benchmarksperformance: avg 11.9% drop, up to 18.9%Hurt
disable early return optimization (ablation)FSA kernel · kernel benchmarksperformance: avg 18.2% drop, up to 25.2%Hurt
separate reduction kernel and online softmax kernelFSA · kernel designmemory access and FLOPs: FSA lower than NSA for g<=8Improved
larger GQA groupFSA vs NSA kernel speedup · kernel latencyspeedup: gap widens for g in {1,2}; gone at g=8Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy