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

• URL: http://arxiv.org/abs/2607.07127v1
• Date: Wed, 08 Jul 2026 08:14:21 GMT
• Title: Weight-Space Physics: Interpretable Hypernetworks for Lattice Quantum Field Theories
• Abstract summary: We introduce JEPAWG, a Joint-Embedding Predictive Architecture-based Weight Generator.<n>It maps couplings directly to flow weights via a learned latent space.<n>As a generator, JEPAWG also interpolates and extrapolates to unseen couplings effectively.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• 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.

Related papers

• Can Deep Learning Achieve Cross-Physics Mapping? [7.103004812157351]
  Cross-Physics Mapping (CPM) is an operator-learning framework for mappings between heterogeneous physical domains.<n>Seven architectures-ResUNet, DeepONet, Fourier, latent, wavelet, U-shaped, and Galerkin neural operators-are evaluated for both diffusion-to-wave and wave-to-diffusion mappings.
  arXiv  Detail & Related papers  (2026-09-15T08:45:23Z)
• PhysiFormer: Learning to Simulate Mechanics in World Space [56.16644631495]
  We present PhysiFormer, a diffusion transformer for physically-plausible 3D object motion.<n>It represents objects as 3D meshes expressed in world coordinates.<n>It substantially outperforms autoregressive baselines in trajectory accuracy, rigidity preservation, and momentum-based physical consistency.
  arXiv  Detail & Related papers  (2026-06-25T17:59:00Z)
• UniPixie: Unified and Probabilistic 3D Physics Learning via Flow Matching [29.8272108619491]
  We introduce UNIPIXIE, a framework trained to predict a continuous path of physically plausible material properties from a single visual input.<n>By learning a direct mapping along an object's softest-to-stiffest spectrum on our PIXIEMULTIVERSE dataset, UNIPIXIE allows for controllable generation of diverse, physically valid material fields.
  arXiv  Detail & Related papers  (2026-06-03T20:08:23Z)
• Mesh Field Theory: Port-Hamiltonian Formulation of Mesh-Based Physics [5.4806374384787695]
  MeshFT-Net is a structure-preserving framework for mesh-based continuum physics.<n>It cleanly separates the physics' topological structure from its metric structure.<n>By eliminating non-physical degrees of freedom and learning only metric-dependent structure, MeshFT provides a principled inductive bias for stable, faithful, and data-efficient learning-based physical simulation.
  arXiv  Detail & Related papers  (2026-05-01T04:35:21Z)
• The Inductive Bias of Convolutional Neural Networks: Locality and Weight Sharing Reshape Implicit Regularization [57.37943479039033]
  We study how architectural inductive bias reshapes the implicit regularization induced by the edge-of-stability phenomenon in gradient descent.<n>We show that locality and weight sharing fundamentally change this picture.
  arXiv  Detail & Related papers  (2026-03-05T04:50:51Z)
• Adiabatic Fine-Tuning of Neural Quantum States Enables Detection of Phase Transitions in Weight Space [0.0]
  We introduce adiabatic fine-tuning, a scheme that trains NQS across a phase diagram, leading to strongly correlated weight representations across different models.<n>This correlation in weight space enables the detection of phase transitions in quantum systems by analyzing the trained network weights alone.<n>Our results establish a connection between physical phase transitions and the geometry of neural network parameters, opening new directions for the interpretability of machine learning models in physics.
  arXiv  Detail & Related papers  (2025-03-21T13:42:11Z)
• Hamiltonian Mechanics of Feature Learning: Bottleneck Structure in Leaky ResNets [58.460298576330835]
  We study Leaky ResNets, which interpolate between ResNets and Fully-Connected nets depending on an 'effective depth'<n>We leverage this intuition to explain the emergence of a bottleneck structure, as observed in previous work.
  arXiv  Detail & Related papers  (2024-05-27T18:15:05Z)
• Emergent Holographic Forces from Tensor Networks and Criticality [0.0]
  We formulate an efficiently implementable multi-scale entanglement renormalization ansatz model of AdS/CFT. We show that the bulk theory arising from this optimized tensor network furnishes excitations with attractive interactions.
  arXiv  Detail & Related papers  (2024-01-24T17:03:03Z)
• End-To-End Latent Variational Diffusion Models for Inverse Problems in High Energy Physics [61.44793171735013]
  We introduce a novel unified architecture, termed latent variation models, which combines the latent learning of cutting-edge generative art approaches with an end-to-end variational framework. Our unified approach achieves a distribution-free distance to the truth of over 20 times less than non-latent state-of-the-art baseline.
  arXiv  Detail & Related papers  (2023-05-17T17:43:10Z)
• Learning Physical Dynamics with Subequivariant Graph Neural Networks [99.41677381754678]
  Graph Neural Networks (GNNs) have become a prevailing tool for learning physical dynamics. Physical laws abide by symmetry, which is a vital inductive bias accounting for model generalization. Our model achieves on average over 3% enhancement in contact prediction accuracy across 8 scenarios on Physion and 2X lower rollout MSE on RigidFall.
  arXiv  Detail & Related papers  (2022-10-13T10:00:30Z)
• Bayesian Neural Networks for Fast SUSY Predictions [58.720142291102135]
  In this paper, machine learning is used to model the mapping from the parameter space of a BSM theory to some of its predictions. All three quantities are modeled with average percent errors of 3.34% or less and in a time significantly shorter than is possible with the supersymmetry codes from which the results are derived. Results are a further demonstration of the potential for machine learning to model accurately the mapping from the high dimensional spaces of BSM theories to their predictions.
  arXiv  Detail & Related papers  (2020-07-09T01:45:06Z)
• Lorentz Group Equivariant Neural Network for Particle Physics [58.56031187968692]
  We present a neural network architecture that is fully equivariant with respect to transformations under the Lorentz group. For classification tasks in particle physics, we demonstrate that such an equivariant architecture leads to drastically simpler models that have relatively few learnable parameters.
  arXiv  Detail & Related papers  (2020-06-08T17:54:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.