Speculative Sampling For Faster Molecular Dynamics
2026-06-01 • Machine Learning
Machine Learning
AI summaryⓘ
The authors introduce Langevin Speculative Dynamics (LSD), a new way to speed up molecular dynamics (MD) simulations, which usually run step-by-step and can be slow. LSD uses a fast, rough model to guess future simulation steps and checks these guesses in parallel with the accurate but slower model, ensuring no extra errors. They show that this method works for different systems, can speed up simulations by 3 to 9 times, and that it correctly follows the intended physics. The approach is inspired by similar speculative techniques used in language and image modeling.
Molecular DynamicsLangevin DynamicsSpeculative SamplingTransport MapModel-AgnosticParallel ComputingSecond-Order DynamicsSampler SpeedupTrajectory Sampling
Authors
Arthur Kosmala, Stephan Günnemann, Meng Gao, Brandon Wood
Abstract
Molecular dynamics (MD) is a key tool for simulating the dynamical behavior of atomic systems. However, MD is inherently serial, which makes it difficult to increase single-system throughput with concurrent compute. To address this, we introduce Langevin Speculative Dynamics (LSD), a distributed and model-agnostic speculative sampler for accelerating MD without adding relative error. Inspired by speculative methods in language and diffusion modeling, LSD uses a draft model to propose fast simulation steps and verifies them in parallel with a slower target model, applying a transport map from the draft to the target distribution. We extend speculative sampling to second-order Langevin dynamics, derive the achievable speedup as a function of physical parameters, show that LSD generalizes across different systems and draft-target combinations with a 3-9x speedup, and confirm theoretically and empirically that LSD samples trajectories from its target model distribution.