AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

SSA: Sparse Sparse Attention by Aligning Full and Sparse Attention Outputs in Feature Space

Zhenyi Shen, Junru Lu, Lin Gui, Jiazheng Li, Yulan He, di yin, Xing Sun

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 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
add alignment loss (bidirectional) to SSA training (ablation)SSA 300M · Commonsense avgaverage accuracy: 48.38 -> 49.69Improved
train with a mix of full and sparse streams (FullRatio=0.5) instead of sparse only (ablation)SSA 300M · Commonsense avgaverage accuracy: 49.00 -> 49.69Improved
mix in the sparse stream (FullRatio=0.5) instead of full stream only (ablation)SSA 300M · Wikitext PPLperplexity: 25.16 -> 24.20Improved
use only one direction of alignment loss (Full to Sparse or Sparse to Full) (ablation)SSA 300M · training stabilityPPL / Comm. Avg.: NaN, 0.00Hurt
sample the attention mode randomly per layer instead of per step (ablation)SSA 300M · training stabilityPPL / Comm. Avg.: NaN, 0.00Hurt
train with larger receptive field (16x64 instead of inference-only extrapolation from 8x16) (ablation)SSA 300M · Commonsense avgaverage accuracy: 49.96 -> 49.16Hurt
train with receptive field 8x16 instead of extrapolating to it (ablation)SSA 300M · Commonsense avgaverage accuracy: 49.69 -> 48.88Hurt
alignment coefficient alpha 10 instead of 5 (ablation)SSA 300M · Commonsense avgaverage accuracy: 48.71 -> 49.69Improved
use L2 alignment loss instead of SmoothL1 (ablation)SSA 300M · Commonsense avgaverage accuracy: 49.69 -> 49.27Hurt
add sliding window to NSA (ablation)NSA RF=256 · Commonsense avgaverage accuracy: 58.9 -> 59.47Improved
add sliding window to NSA (ablation)NSA, PG-19 32K RoPE scaling · PG-19 PPL 32Kperplexity: 18.16 -> 15.00Improved
SSA training instead of MoBA at sparse inference RF=2561B model · LongBench 32k extrapolationLongBench score: 16.88 -> 17.89Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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