Generative Adaptation of Dynamics to Environmental Shifts via Weight-space Diffusion

TL;DR

DynaDiff is a generative meta-learning framework that enables rapid adaptation of dynamics models to environmental shifts by generating weights directly, improving accuracy and efficiency over traditional methods.

Contribution

It introduces a novel weight-space diffusion approach with weight graphs and a dynamics-informed prompter for fast, data-efficient model adaptation.

Findings

Boosts average prediction accuracy by 10.78% over baselines.

Pre-constructs a model zoo to amortize fine-tuning costs.

Achieves significant deployment efficiency improvements.

Abstract

Data-driven dynamics prediction often fails under environmental shifts, while traditional fine-tuning remains computationally prohibitive for hardware-constrained or data-scarce applications. We propose DynaDiff, a generative meta-learning framework that transitions the paradigm from gradient-based tuning or modulation to direct weight-space generation. Specifically, we first abstract expert weights as novel weight graphs, utilizing multi-head attention to explicitly capture topological coupling within weights. Subsequently, we design a functional loss to ensure that the generated models achieve consistency with expert models in physical behavior. Finally, we develop a dynamics-informed prompter that extracts cross-domain physical and spectral features from observation sequences to condition the diffusion model. Experiments demonstrate that DynaDiff boosts average prediction accuracy by 10.78% over competitive baselines. Furthermore, by pre-constructing a model zoo of expert predictors, we amortize the fine-tuning overhead into a one-time offline cost, significantly boosting deployment efficiency in new environments.