Learning 3D Anisotropic Noise Distributions Improves Molecular Force Field Modeling

Coordinate denoising has emerged as a promising method for 3D molecular pre-training due to its theoretical connection to learning a molecular force field. However, existing denoising methods rely on oversimplified molecular dynamics that assume atomic motions to be isotropic and homoscedastic. To address these limitations, we propose a novel denoising framework AniDS: sotropic Variational Autoencoder for 3D Molecular enoi ing. AniDS introduces a structure-aware anisotropic noise generator that can produce atom-specific, full covariance matrices for Gaussian noise distributions to better reflect directional and structural variability in molecular systems. These covariances are derived from pairwise atomic interactions as anisotropic corrections to an isotropic base. Our design ensures that the resulting covariance matrices are symmetric, positive semi-definite, and SO(3)-equivariant, while providing greater capacity to model complex molecular dynamics. Extensive experiments show that AniDS outperforms prior isotropic and homoscedastic denoising models and other leading methods on the MD17 and OC22 benchmarks, achieving average relative improvements of 8.9\