AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Less Tokens, Better Forecasts: Sparse Residual Routing for Efficient Weather Prediction

Janet Wang, Yunbei Zhang, Lin Zhao, Xi Xiao, Jihun Hamm, Xiao Wang

ECCV 2026 · Architectures and efficiency › Attention and transformers · approach: Block-sparse attention kernels · 1 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 Sparse-Reslim routing (r=0.25, split 2,8,2)Res-Slim-ViT · ERA5 1.0° Z500RMSE: 868.7 -> 768.3Improved
add Sparse-Reslim routingRes-Slim-ViT · ERA5 1.0° T2mRMSE: 3.19 -> 2.68Improved
add Sparse-Reslim routingRes-Slim-ViT · ERA5 0.25° Z500RMSE: 1042 -> 892Improved
add Sparse-Reslim routing to EDM generative modelEDM ViT · ERA5 1.0° Z500RMSE: 951.3 -> 889.7Improved
add Sparse-Reslim routing to EDM generative modelEDM ViT · ERA5 1.0°training throughput (samples/s/GPU): 11.4 -> 20.8Improved
add Sparse-Reslim routingRes-Slim-ViT · ERA5 1.40625° T2m 10-day rolloutACC: 0.53 -> 0.69Improved
lower keep ratio from 0.25 to 0.10 (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 801.5Hurt
raise keep ratio to 1.0 (dense) (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 868.7Hurt
use one dense early and late block (1,10,1) instead of (2,8,2) (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 812.4Hurt
use four dense blocks per end (4,4,4) instead of (2,8,2) (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 831.2Hurt
replace random selection with top-K attention score (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 784.1Hurt
replace random selection with learned difficulty head (ablation)Sparse-Reslim · ERA5 1.0° Z500RMSE: 768.3 -> 779.8Hurt

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