AI RESEARCH
A recipe for scalable attention-based MLIPs: unlocking long-range accuracy with all-to-all node attention
arXiv CS.LG
•
ArXi:2603.06567v1 Announce Type: new Machine-learning interatomic potentials (MLIPs) have advanced rapidly, with many top models relying on strong physics-based inductive biases. However, as models scale to larger systems like biomolecules and electrolytes, they struggle to accurately capture long-range (LR) interactions, leading current approaches to rely on explicit physics-based terms or components. In this work, we propose AllScAIP, a straightforward, attention-based, and energy-conserving MLIP model that scales to O(100M)