AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

(Sparse) Attention to the Details: Preserving Spectral Fidelity in ML-based Weather Forecasting Models

Maksim Zhdanov, Ana Lucic, Max Welling, Jan-Willem van de Meent

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Long-context KV and memory 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
compress latent grid (enter at Nside=32 instead of 64) (ablation)MOSAIC · 10m wind spectrum, ERA5 1.5degspectral ratio near Nyquist: 0.97 -> 1.02Hurt
predict residual x_{t+1}-x_t instead of next state directly (ablation)MOSAIC · 10m wind spectrum, ERA5 1.5degspectral ratio at highest wavenumber: 0.97 -> 1.00Hurt
train probabilistically with ensembles instead of deterministically to the ensemble meanML weather models · 10m wind spectrum, 1.5degspectral ratio at highest wavenumber: 0.24 (Stormer) -> 0.97 (MOSAIC)Improved
use block-sparse attention instead of dense FlashAttention and NSAattention layer · forward pass runtime up to 512k tokensspeedup: 61.8x over dense, 9.4x over NSAImproved
inject noise z as additive bias in SwiGLU gates instead of conditioning layer norm (ablation)functional perturbations · ensemble calibrationspread-to-skill ratio: 0.83 to 0.95Improved
MOSAIC (1.5deg) vs 0.25deg-trained deterministic modelsMOSAIC · Z500 10-day RMSERMSE: 610 vs Aurora 668, Pangu 800, GraphCast 748Improved
MOSAIC vs ArchesWeatherGen on Z500 10-day ensemble meanMOSAIC · ERA5 1.5deg Z500RMSE: 610.36 -> 624.08Hurt

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