AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Elastic Attention: Test-time Adaptive Sparsity Ratios for Efficient Transformers

Zecheng Tang, Quantong Qiu, Yi Yang, Zhiyi Hong, Haiya Xiang, Kebin Liu, Qingqing Dang, Juntao Li, Min Zhang

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs · 0 citations (OpenAlex)

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
enlarge router MLP hidden size from 2x to 4x d' (ablation)Elastic Attention (Qwen3-4B) · Avg of LongBench, RULER, LongBench-V2score: 45.22 -> 45.92Improved
enlarge router MLP hidden size from 4x to 8x d' (ablation)Elastic Attention (Qwen3-4B) · Avg of LongBench, RULER, LongBench-V2score: 45.92 -> 46.40Mixed
use XAttention for retrieval heads too (XA-SSA, full sparse regime) (ablation)Qwen3-4B · Avg of LongBench, RULER, LongBench-V2score: 46.80 -> 45.93Hurt
use XAttention for retrieval heads too (XA-SSA) (ablation)Qwen3-8B · Avg of LongBench, RULER, LongBench-V2score: 53.26 -> 49.42Hurt
use XAttention for retrieval heads too (XA-SSA) (ablation)Llama-3.1-8B-Instruct · Avg of LongBench, RULER, LongBench-V2score: 56.48 -> 50.02Hurt
use XAttention instead of full attention for retrieval heads (FA-XA vs FA-SSA sparse heads)Elastic Attention on Llama-3.1-8B-Instruct · RULER averageaccuracy: 72.85 -> 81.82Improved
use XAttention sparse heads instead of streaming sparse headsElastic Attention on Qwen3-8B · LongBench-V2accuracy: 33.41 -> 30.74Hurt
lower target sparsity t_sen for sensitive tasks from 0.7 to 0.4 (ablation)Elastic Attention · overall long-context performanceperformance: t_sen 0.7 -> 0.4Improved
replace retrieval heads with sparse heads, Omega_MSR 0.0 to 0.2 (ablation)Llama-3.1-8B-Instruct (static head replacement) · LongBench Single-Doc QAretention %: 100.00 -> 71.32Hurt
replace retrieval heads with sparse heads, Omega_MSR 0.0 to 0.2 (ablation)Llama-3.1-8B-Instruct (static head replacement) · LongBench Summarizationretention %: 100.00 -> 98.00No clear effect
raise Omega_MSR from 0.0 to 1.0 by replacing retrieval heads (ablation)Llama-3.1-8B-Instruct (static head replacement) · LongBench Coderetention %: 100.00 -> 99.46No clear effect
Elastic Attention vs PruLong at fixed 0.70 sparsityLlama-3.1-8B-Instruct · RULER averageaccuracy: 48.82 -> 72.85Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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