Newest papers

Action Learning for Scientific Simulations

Sarah Liaw, Yuanqi Du, Yilun Du · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31585 · cited by 0 works (OpenAlex) · open-access PDF

French

La compréhension et le contrôle de l'objectivation de la façon dont les systèmes évoluent sont au cœur de la science et de l'ingénierie, où de nombreux problèmes nécessitent un raisonnement sur les trajectoires complètes. Les approches existantes apprennent généralement la dynamique locale ou génèrent des trajectoires, mais les objectifs construits à partir de transitions individuelles peuvent manquer des dépendances au fil du temps, tandis que les modèles génératifs ne fournissent pas nécessairement un objectif traçable pour évaluer et optimiser les chemins arbitraires. Nous étudions $\textit{action learning}$, le problème de la récupération d'une action stochastique à partir de trajectoires observations, donnant un objectif commun pour le raisonnement sur les systèmes dynamiques

Machine translation (Meta M2M100); the English original follows.

Abstract

Understanding and controlling how systems evolve is central to science and engineering, where many problems require reasoning about complete trajectories. Existing approaches typically learn local dynamics or generate trajectories, but objectives constructed from individual transitions can miss dependencies across time, while generative models do not necessarily provide a tractable objective for evaluating and optimizing arbitrary paths. We study $\textit{action learning}$, the problem of recovering a stochastic action from trajectory observations, yielding a common objective for reasoning about dynamical systems. We introduce $\textbf{LEAP}$ ($\textbf{L}$earned $\textbf{E}$nergy on $\textbf{A}$ction $\textbf{P}$aths), which anchors the unknown action to a tractable reference process and learns the remaining correction by comparing observed trajectories with reference trajectories. This construction yields a differentiable objective on path space that can capture both local dynamical structure and nonadditive dependencies across time. Once learned, the same action can be flexibly reused for path ranking and inference, mechanism steering, constrained optimization and sampling. Across stochastic physical systems, many-body orbital dynamics and molecular simulations, LEAP recovers underlying dynamics, captures dependencies across time and enables transfer across inference objectives and physical conditions. We show that $\textit{action learning}$ is a promising framework for turning trajectory observations into reusable objectives for scientific inference and control.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Inference, Trajectory, Action (physics), Path (computing), Process (computing)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange