MetaSym: A Symplectic Meta-learning Framework for Physical Intelligence

TL;DR

MetaSym is a novel deep learning framework that incorporates symplectic inductive biases to ensure physical invariants, enabling efficient adaptation and superior performance across diverse physics-based tasks, including real-world robotics data.

Contribution

MetaSym introduces a symplectic encoder and meta-attention-based autoregressive decoder, combining physical invariants with flexible adaptation for physics-aware deep learning.

Findings

Outperforms state-of-the-art models in few-shot learning.

Demonstrates robustness to noise and real-world uncertainties.

Effective across diverse systems including quantum and robotic dynamics.

Abstract

Scalable and generalizable physics-aware deep learning has long been considered a significant challenge with various applications across diverse domains ranging from robotics to molecular dynamics. Central to almost all physical systems are symplectic forms, the geometric backbone that underpins fundamental invariants like energy and momentum. In this work, we introduce a novel deep learning framework, MetaSym. In particular, MetaSym combines a strong symplectic inductive bias obtained from a symplectic encoder, and an autoregressive decoder with meta-attention. This principled design ensures that core physical invariants remain intact, while allowing flexible, data efficient adaptation to system heterogeneities. We benchmark MetaSym with highly varied and realistic datasets, such as a high-dimensional spring-mesh system Otness et al. (2021), an open quantum system with dissipation and measurement backaction, and robotics-inspired quadrotor dynamics. Crucially, we fine-tune and deploy MetaSym on real-world quadrotor data, demonstrating robustness to sensor noise and real-world uncertainty. Across all tasks, MetaSym achieves superior few-shot adaptation and outperforms larger state-of-the-art (SOTA) models.