AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

A recipe for scalable attention-based ML potentials: unlocking long-range accuracy with all-to-all node attention

Eric Qu, Brandon M. Wood, Aditi S. Krishnapriyan, Zachary Ward Ulissi

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer neural operators for PDEs

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 LAE angular encoding (ablation)AllScAIP-md 85M · OMol25 4M Totalforce MAE: 0.007603682161 -> 0.006889121209Improved
add ERoPE distance bias (ablation)AllScAIP-md 85M · OMol25 4M Totalenergy MAE: 0.001293416426 -> 0.001164967593Improved
add all-to-all node attention (ablation)AllScAIP-md 85M · OMol25 4M Totalenergy MAE: 0.001563449868 -> 0.001164967593Improved
add all-to-all node attention (ablation)AllScAIP-md 85M · OMol25 102M Totalforce MAE: 0.006148287564 -> 0.005096293718Improved
add LAE (ablation)AllScAIP-md 85M · OMol25 102M Totalenergy MAE: 0.0007254301528 -> 0.0006647585806No clear effect
add LAE and ERoPE (ablation)AllScAIP-md 85M · OMol25 102M Totalenergy MAE: 0.0006407190255 (without) vs 0.0006647585806 (with)No clear effect
scale data from 4M to 102MAllScAIP-md 85M · OMol25 Totalforce MAE: 0.006889121209 -> 0.005096293718Improved
add rotation-equivariance by learning (train on OMol-102M vs random init)AllScAIP-md · Random rotation force cosine similaritycosine similarity: 0.6827 -> 0.9999Improved
replace direct-force with conservative fine-tuneAllScAIP-md · OMol25 4M Totalenergy MAE: 0.90 (ft-cons., 4M)Improved
use all-to-all node attention vs eSEN/UMAAllScAIP-md-cons. · OMol25 distance scaling (LR)energy error: about 90% reduction vs second bestImproved

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