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 $\nu \approx 1$, 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.

We present results for the chromo-electric field generated by a static quark-antiquark pair at nearly zero temperature in lattice QCD with 2+1 dynamical staggered fermions at physical quark masses. We investigate the evolution of the flux-tube structure as the distance between the static color charges increases. We find hints that string breaking occurs at a distance in the range $0.963 \; \text{fm} \; \lesssim \; d^* \lesssim \; 1.156 \; \text{fm}$.

We use the step-scaling method and results obtained at $N = 3, 5$ and $8$ to determine the $N$-dependence of the dynamically generated scale $\Lambda$ of $\mathrm{SU}(N)$ Yang--Mills theories. We implement the step-scaling method in a suitable finite-volume renormalization scheme based on twisted boundary conditions, introduced to effectively achieve large-$N$ volume independence, and on a coupling defined through the gradient flow. In the $\overline{\mathrm{MS}}$ scheme, we obtain the following values in terms of the gradient flow scale $t_0$: $\sqrt{8t_0}\Lambda_{\scriptscriptstyle{\overline{\mathrm{MS}}}} = 0.577(23)$, $0.632(32)$, and $0.611(43)$ for $N=3,5$ and $8$, respectively. They extrapolate to a large-$N$ value of: $\sqrt{8t_0}\Lambda_{\scriptscriptstyle{\overline{\mathrm{MS}}}} (N=\infty) = 0.639(36)$, and the $N$-dependence is given by $\sqrt{8t_0}\Lambda_{\scriptscriptstyle{\overline{\mathrm{MS}}}}(N)=0.639(36)[1-0.85(62)/N^2+\mathcal{O}(1/N^4)]$. This work represents the first calculation of the Yang--Mills $\Lambda$-parameter in the large-$N$ limit that does not rely on asymptotic scaling strategies.

We calculate the full set of BSM contributions to the inclusive non-leptonic $B$-meson lifetime from charmless final states within the framework of the Weak Effective Theory up to next-to-leading order in QCD and to leading power in the heavy-quark expansion. We do so by computing cut diagrams and the corresponding phase space integrals. This involves calculating current-current, penguin-penguin, penguin-dipole and dipole-dipole diagrams with two-, three- and four-particle cuts. We describe the technical difficulties of the computation. We then discuss how our results can be used to constrain new physics scenarios related to the anomalies observed in semi-leptonic and charmless non-leptonic decays.

The deconfined quantum critical point (DQCP) in square lattice S=1/2 quantum antiferromagnets has been extensively studied with a large body of evidence pointing to a weakly first-order transition scenario. Recent studies, which focused on entanglement at this nearly continuous DQCP in square lattice J-Q models, have observed conflicting bipartite entanglement entropy (EE) scaling behavior. One bipartition choice gave scaling coefficients in remarkable agreement with predictions from the unitary CFT corresponding to the putative DQCP. While another equally natural choice gave scaling coefficients in complete violation of unitary CFT that may be attributed to lack of scale invariance at the known weakly first-order behavior of the model. This motivates the exploration of DQCP behavior via entanglement measures in lattice models with distinct crystalline symmetries. Here we study a S=1/2 honeycomb model that hosts a nearly continuous transition between Néel and valence-bond-solid ground states relevant to probing DQCP. Using large-scale quantum Monte Carlo simulations, we compute the Rényi EE for a variety of bipartitions and test the CFT based description of the DQCP on the honeycomb lattice. For smooth bipartitions, we find no evidence of logarithmic corrections, in accordance with CFT, thereby essentially ruling out contributions from Goldstone modes. For subsystems with corners, CFT predicts universal logarithmic contributions, which we extract for corners with 60 and 120 degree angles and find close agreement with an emergent SO(5) CFT. While we observe scaling consistent with a critical system in the majority of cases, we also demonstrate an intriguing counterexample of the hexagon subsystem that exhibits a subtle period three oscillation. This results in three separate finite-size series, where the sign of the logarithmic term apparently changes depending on the series.