Scale-Aware Adversarial Analysis: A Diagnostic for Generative AI in Multiscale Complex Systems

2026-05-01

AI Summary

It remains unclear whether current generative models genuinely learn the intrinsic dynamics of multiscale physical systems or merely capture statistical correlations, due to a lack of effective evaluation methods. Conventional explainable AI techniques often introduce non-physical artifacts through pixel-level perturbations, hindering assessments of physical consistency. This work proposes the Constrained Diffusion Decomposition (CDD) diagnostic framework, which applies physically constrained, deterministic interventions to Denoising Diffusion Probabilistic Models (DDPMs) in a continuous multiscale space. CDD uniquely integrates multiscale physical priors into generative model evaluation by constructing a physically consistent continuous state space, enabling controlled tests of model causality and cross-scale stability. Experiments reveal that unconstrained models exhibit structural freezing and nonlinear instabilities under moderate physical perturbations, failing to preserve the multiscale evolutionary characteristics of real physical systems.

Abstract

Complex physical systems, from supersonic turbulence to the macroscopic structure of the universe, are governed by continuous multiscale dynamics. While modern machine learning architectures excel at mapping the high-dimensional observables of these systems, it remains unclear whether they internalize the governing physical laws or merely interpolate discrete statistical correlations. Standard Explainable AI (XAI) architectures, particularly perturbation-based and gradient-saliency methods, rely on pixel-wise perturbations, which generate unphysical artifacts and push inputs off the valid empirical distribution. To resolve this, we introduce a diagnostic framework driven by Constrained Diffusion Decomposition (CDD), a diffusion-based multiscale data decomposition algorithm that enables physically constrained data generation and model evaluation via scale-aware modifications. Applying this framework to a Denoising Diffusion Probabilistic Model (DDPM), we execute deterministic interventions directly within the continuous, CDD-based scale space. We demonstrate that under moderate physical perturbations, the unconstrained generative model exhibits localized structural freezing and non-linear instability rather than continuous PDE-like responses. The network fails to maintain cross-scale continuity, causing the generative trajectory to diverge when pushed into unseen physical states. By synthesizing a continuum of physically coherent states, this scale-informed methodology establishes a controlled test ground to evaluate algorithmic vulnerabilities, providing the rigorous physical constraints necessary for future architectures to respect the multiscale causality of the natural universe.

Problem

• multiscale complex systems
• generative AI
• physical consistency
• adversarial analysis
• explainable AI

Innovation

• Constrained Diffusion Decomposition
• Scale-Aware Adversarial Analysis
• Multiscale Generative Modeling
• Physically Consistent Perturbations
• Denoising Diffusion Probabilistic Model