AtomWorld-Mem: Memory-Restored World States for Long-Horizon Atomistic Evolution

2026-09-25

AI Summary

This study addresses the challenge of long-term evolution prediction arising from the ambiguity of instantaneous atomic snapshots, which fail to capture latent dynamical contexts. To this end, we propose a memory-retrieval world model that reformulates atomic evolution as a memory recovery problem. The method employs a spatial encoder and a long short-term memory network to integrate multi-scale keyframes for reconstructing latent world states, combined with kinetic Monte Carlo (KMC)-constrained prioritized sampling to identify physically valid events. Experimental results demonstrate that, under a fixed event budget, the proposed model significantly enhances both the efficiency and fidelity of long-term evolution predictions. Furthermore, it exhibits excellent zero-shot generalization and transferability across diverse alloy systems.

Abstract

High-fidelity atomistic evolution over long timescales requires more than observing the current crystal configuration. Instantaneous atomistic snapshots are often incomplete: locally similar configurations can correspond to different hidden dynamical contexts, future event preferences, and waiting-time scales. We argue that this snapshot ambiguity makes long-horizon atomistic evolution fundamentally a memory-based world-state restoration problem. To address this, we introduce AtomWorld-Mem, a memory-restored atomistic world model that recovers the latent world state missing from instantaneous crystal snapshots. AtomWorld-Mem treats the evolving alloy as an AtomWorld: spatial encoders write multi-scale atomistic keyframes from dense local topology and sparse long-range defect context, while short-term event memory and long-term structural memory integrate these keyframes across time to restore a future-predictive evolutionary state. The restored state is used to prioritize legal vacancy-mediated events under single-event Kinetic Monte Carlo (KMC) constraints, while event legality, physical execution, and residence-time updates remain governed by the underlying simulator. Empirically, AtomWorld-Mem improves long-horizon atomistic progress under fixed microscopic event budgets while maintaining high-fidelity evolution across energetic, structural, and vacancy-transport observables. It further transfers zero-shot across diverse unseen alloy-temperature AtomWorlds, suggesting that the learned memory-restoration mechanism captures reusable principles of hidden-state inference rather than a system-specific local energy heuristic. These results position memory-restored world-state modeling as a promising route toward efficient, physically grounded, and transferable atomistic evolution.

Problem

• atomistic evolution
• long-horizon simulation
• snapshot ambiguity
• world-state restoration
• hidden dynamics

Innovation

• Memory-Restored World Model
• Atomistic Evolution
• Kinetic Monte Carlo
• Zero-Shot Transfer
• Latent State Inference