LLM4Fluid: Large Language Models as Generalizable Neural Solvers for Fluid Dynamics

2026-01-29

AI Summary

This work proposes LLM4Fluid, a novel framework that leverages pre-trained large language models as universal neural solvers for fluid dynamics. Addressing the limited generalization of conventional deep learning approaches—which typically require retraining for new flow conditions—LLM4Fluid introduces a physics-informed, decoupled reduced-order modeling strategy to map high-dimensional flow fields into a latent space. By integrating modal alignment with an autoregressive temporal prediction mechanism, the framework enables accurate, long-term forecasting of flow evolution across diverse scenarios without any retraining. Demonstrating zero-shot and in-context learning capabilities, LLM4Fluid achieves high predictive accuracy on a range of complex flows and significantly outperforms existing methods, highlighting its exceptional generalization performance.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Deep learning has emerged as a promising paradigm for spatio-temporal modeling of fluid dynamics. However, existing approaches often suffer from limited generalization to unseen flow conditions and typically require retraining when applied to new scenarios. In this paper, we present LLM4Fluid, a spatio-temporal prediction framework that leverages Large Language Models (LLMs) as generalizable neural solvers for fluid dynamics. The framework first compresses high-dimensional flow fields into a compact latent space via reduced-order modeling enhanced with a physics-informed disentanglement mechanism, effectively mitigating spatial feature entanglement while preserving essential flow structures. A pretrained LLM then serves as a temporal processor, autoregressively predicting the dynamics of physical sequences with time series prompts. To bridge the modality gap between prompts and physical sequences, which can otherwise degrade prediction accuracy, we propose a dedicated modality alignment strategy that resolves representational mismatch and stabilizes long-term prediction. Extensive experiments across diverse flow scenarios demonstrate that LLM4Fluid functions as a robust and generalizable neural solver without retraining, achieving state-of-the-art accuracy while exhibiting powerful zero-shot and in-context learning capabilities. Code and datasets are publicly available at https://github.com/qisongxiao/LLM4Fluid.

Problem

• fluid dynamics
• generalization
• spatio-temporal modeling
• neural solvers
• zero-shot learning

Innovation

• Large Language Models
• Fluid Dynamics
• Reduced-Order Modeling
• Modality Alignment
• Zero-shot Learning