Weight-Space Physics: Interpretable Hypernetworks for Lattice Quantum Field Theories

2026-07-08

AI Summary

This work addresses the opaque relationship between coupling constants and physical observables in traditional lattice field theories, as well as the limited interpretability of existing normalizing flow approaches. To overcome these challenges, the authors propose the JEPAWG architecture, which leverages lattice field theory as a testbed for neural network interpretability. By employing a Joint Embedding Predictive Architecture (JEPA)-driven hypernetwork, the method directly generates normalizing flow weights from coupling parameters, treating the network weights themselves as novel physical observables. The approach successfully recovers the intrinsic dimensionality of the theoretical manifold in weight space, accurately locates the phase transition belonging to the 2D Ising universality class, and encodes finite-size scaling behavior. Furthermore, it achieves robust interpolation and extrapolation to unseen coupling parameters across scalar field theories on lattices ranging from \(6^2\) to \(11^2\), significantly outperforming baseline methods such as PCA, autoencoders, and variational autoencoders.

Abstract

Lattice field theory is the workhorse of non-perturbative physics, used to simulate phenomena from the strong nuclear force to critical phenomena in materials. Its Boltzmann distributions are parametrized analytically by coupling constants, but these bare parameters are weak predictors of observables -- extracting physics typically requires extensive simulation. While normalizing flows have emerged as effective samplers at fixed couplings, it remains difficult to interpret what these networks have learned. This raises a natural question: can the physics be read off directly from the flow network parameters themselves, and can those parameters be generated for unseen theories? We propose lattice field theory as a testbed for neural network interpretability: because the target physics is qualitatively well-understood and smoothly varying, it provides ideal synthetic data with known ground truth. To this end, we introduce JEPAWG, a Joint-Embedding Predictive Architecture-based Weight Generator that maps couplings directly to flow weights via a learned latent space. On a scalar theory at lattices of size $6^2$ to $11^2$, the JEPAWG latent space recovers the correct intrinsic dimension of the underlying manifold, locates the phase transition, and encodes a finite-size shift aligned with the 2D Ising exponent $ν\approx 1$, allowing us to uncover physical structure by studying the network weights alone. This suggests the fascinating idea of treating the network weights as a new type of physical observable. As a generator, JEPAWG also interpolates and extrapolates to unseen couplings effectively and remains robust to weight-space discontinuities introduced by multi-seed training data, outperforming PCA, AE, and VAE baselines.

Problem

• lattice field theory
• neural network interpretability
• hypernetworks
• physical observables
• weight-space

Innovation

• hypernetworks
• lattice field theory
• neural interpretability
• normalizing flows
• phase transition