{"slug": "least-action-guided-diffusion-for-physical-extrapolation", "title": "Least-Action-Guided Diffusion for Physical Extrapolation", "summary": "Researchers have developed a least-action-guided diffusion (LAPG) framework that enforces physical consistency in generative models during inference, addressing the challenge of reliable extrapolation beyond training data. The method combines a conditional score-based diffusion model with an action-derived physical guidance score, reducing phase drift and preserving dissipative decay in tests on ordinary and partial differential equation systems. This approach offers an alternative to pointwise residual penalties and improves predictions for out-of-distribution conditions in computational physics.", "body_md": "arXiv:2606.11277v1 Announce Type: new\nAbstract: Reliable extrapolation remains a central challenge for generative models in computational physics, because models trained over finite ranges of time, parameters, or geometries may produce physically inconsistent predictions outside the training distribution. We introduce a least-action-principle-guided diffusion, LAPG, a framework that promotes physical consistency during inference rather than relying solely on constraints imposed during training. The method combines a conditional score-based diffusion model with an action-derived physical guidance score. In the first stage, the learned score model generates an in-distribution proposal; in the second, an action-based variational prior refines this proposal toward the target out-of-distribution condition. This formulation turns the principle of least action into a differentiable inference-time correction mechanism and provides an alternative to pointwise residual penalties that often require empirical loss balancing.\nWe evaluate LAPG on representative ordinary- and partial-differential-equation systems, including free fall, conservative and dissipative spring-mass dynamics, interacting point vortices, and potential flow over parameterized airfoils. In temporal, parameter, and geometric extrapolation tests, LAPG reduces phase drift, preserves dissipative decay, captures vortex motion, and improves the lift response of airfoil flows compared with training-time physics-informed baselines.", "url": "https://wpnews.pro/news/least-action-guided-diffusion-for-physical-extrapolation", "canonical_source": "https://arxiv.org/abs/2606.11277", "published_at": "2026-06-12 04:00:00+00:00", "updated_at": "2026-06-12 04:03:49.701921+00:00", "lang": "en", "topics": ["machine-learning", "generative-ai", "artificial-intelligence", "neural-networks", "ai-research"], "entities": ["LAPG"], "alternates": {"html": "https://wpnews.pro/news/least-action-guided-diffusion-for-physical-extrapolation", "markdown": "https://wpnews.pro/news/least-action-guided-diffusion-for-physical-extrapolation.md", "text": "https://wpnews.pro/news/least-action-guided-diffusion-for-physical-extrapolation.txt", "jsonld": "https://wpnews.pro/news/least-action-guided-diffusion-for-physical-extrapolation.jsonld"}}