AI models of unstable flow exhibit hallucination

TL;DR

This paper presents evidence of hallucinations in AI fluid dynamics models, identifies their spectral bias origin, and introduces DeepFingers, a framework that improves physical accuracy in modeling viscous fingering phenomena.

Contribution

It is the first to systematically demonstrate hallucinations in AI fluid models, analyze their spectral bias origin, and propose DeepFingers for more accurate, physically consistent predictions.

Findings

AI models exhibit hallucinations like spurious interfaces and reverse diffusion.

Spectral bias causes hallucinations, especially at high flow rates and viscosity contrasts.

DeepFingers effectively captures complex viscous fingering patterns while maintaining physical laws.

Abstract

We report the first systematic evidence of hallucination in AI models of fluid dynamics, demonstrated in the canonical problem of hydrodynamically unstable transport known as viscous fingering. AI-based modeling of flow with instabilities remains challenging because rapidly evolving, multiscale fingering patterns are difficult to resolve accurately. We identify solutions that appear visually realistic yet are physically implausible, analogous to hallucinations in large language models. These hallucinations manifest as spurious fluid interfaces and reverse diffusion that violate conservation laws. We show that their origin lies in the spectral bias of AI models, which becomes dominant at high flow rates and viscosity contrasts. Guided by this insight, we introduce DeepFingers, a new framework for AI-driven fluid dynamics that enforces balanced learning across the full spectrum of spatial modes by combining the Fourier Neural Operator with a Deep Operator Network to predict the spatiotemporal evolution of viscous fingers. By conditioning on both time and viscosity contrast, DeepFingers learns mappings between successive concentration fields across regimes. The framework accurately captures tip splitting, finger merging, and channel formation while preserving global metrics of mixing. The results open a new research direction to investigate fundamental limitations in AI models of physical systems.