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

📄 arXiv: 2607.07127v1 📥 PDF

作者: Tobias Göbel, Julian R. Ebelt, Zier Mensch, Mathis Gerdes, Miranda C. N. Cheng

分类: hep-lat, cs.LG

发布日期: 2026-07-08

备注: 9 + 13 pages, 4 + 8 figures, 3 + 5 tables


💡 一句话要点

提出JEPAWG以解决量子场论中的可解释性问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)

关键词: 量子场论 晶格场论 可解释性 神经网络 物理可观测量 相变 潜在空间

📋 核心要点

  1. 现有方法在提取晶格场论中的物理信息时,依赖于大量模拟,导致效率低下。
  2. 本文提出JEPAWG,通过学习的潜在空间将耦合常数直接映射到流网络权重,提升可解释性。
  3. JEPAWG在标量理论上表现优异,能够有效定位相变并超越传统方法,如PCA和VAE。

📝 摘要(中文)

晶格场论是非微扰物理学的重要工具,用于模拟强核力等现象。现有的博尔兹曼分布通过耦合常数进行参数化,但这些参数对可观测量的预测能力较弱,提取物理信息通常需要大量模拟。本文提出JEPAWG,一种基于联合嵌入预测架构的权重生成器,能够直接将耦合映射到流网络权重,并通过学习的潜在空间有效地处理未见理论。JEPAWG在不同大小的标量理论上表现出色,能够恢复底层流形的内在维度,定位相变,并揭示物理结构,表明网络权重可以作为新的物理可观测量。

🔬 方法详解

问题定义:本文旨在解决晶格场论中物理信息提取的可解释性问题。现有方法在固定耦合下的模拟效率低,且难以解释网络学习到的物理特征。

核心思路:提出JEPAWG,通过学习的潜在空间将耦合常数映射到流网络权重,直接从网络参数中提取物理信息,提供更好的可解释性。

技术框架:JEPAWG的整体架构包括耦合常数的输入、潜在空间的学习、权重生成和物理特征的提取四个主要模块。通过这些模块,网络能够有效地处理未见的耦合并进行插值和外推。

关键创新:JEPAWG的最大创新在于将网络权重视为新的物理可观测量,能够直接反映底层物理结构,与传统方法相比,提供了更高的可解释性和准确性。

关键设计:在设计中,JEPAWG采用了联合嵌入预测架构,使用特定的损失函数来优化权重生成过程,并通过多种训练数据的多种种子保持对权重空间不连续性的鲁棒性。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

JEPAWG在标量理论的实验中表现出色,成功定位相变并恢复底层流形的内在维度。与PCA、AE和VAE等基线相比,JEPAWG在处理未见耦合时表现出更高的鲁棒性和准确性,显示出显著的性能提升。

🎯 应用场景

该研究的潜在应用领域包括量子场论的模拟、材料科学中的相变研究以及其他需要高效提取物理信息的领域。JEPAWG的可解释性和生成能力为未来的物理研究提供了新的工具,可能推动相关领域的进展。

📄 摘要(原文)

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.