AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

NESTOR: A Nested MOE-based Neural Operator for Large-Scale PDE Pre-Training

Dengdi Sun, Xiaoya Zhou, Xiao Wang, Hao Si, Wanli Lyu, Jin Tang, Bin Luo

CVPR 2026 · Architectures and efficiency › Mixture of experts · approach: Sparse MoE as regularization and attention · 2 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 token-level Sub-MoE (ablation)NESTOR · PDEBench 6 sub-tasksAvg L2RE: 0.0197 -> 0.0173Improved
add load balancing loss (ablation)NESTOR · PDEBench 6 sub-tasksAvg L2RE: 0.0178 -> 0.0173Mixed
fuse AFNO and FlashAttention via MoE instead of simple addition (ablation)NESTOR · PDEBench 6 sub-tasksAvg L2RE: 0.0196 -> 0.0173Improved
increase non-shared experts from 6 to 12 (ablation)NESTOR · FT-500 averageAvg L2RE: 0.0213 -> 0.0230Hurt
increase non-shared experts from 2 to 6 (ablation)NESTOR · FT-500 averageAvg L2RE: 0.0222 -> 0.0213Improved
increase non-shared experts from 6 to 12 at FT-200 (ablation)NESTOR · FT-200 averageAvg L2RE: 0.0246 -> 0.0832Hurt
increase pre-training datasets from 3 to 12 (ablation)NESTOR · FT-500 averageAvg L2RE: 0.0234 -> 0.0208Improved
fine-tune 500 epochs instead of 200 (ablation)NESTOR · PDEBench CNS eta=1 avgL2RE: 0.0161 -> 0.0114Improved
NESTOR instead of DPOT-T at pre-trainingPDE pre-training · PDEBench CNS eta=0.1 avgL2RE: 0.0176 -> 0.0161Improved
NESTOR instead of DPOT-T at pre-trainingPDE pre-training · PDEArena NSL2RE: 0.125 -> 0.132Hurt
fine-tune from pre-trained weights instead of training from scratch2D turbulence 512x512 · 2D high-resolution turbulenceL2RE: 47.3% improvementImproved

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