Searcharxiv⌕ Search

arXiv subjects

Joseph Smith

Publications and source records attributed to Joseph Smith.

15 recordsLinked to original sources
arXiv2026

Quantum Effects in Supersymmetric Galilean Yang-Mills

We study the quantum dynamics of supersymmetric Galilean Yang-Mills (SGYM), with a focus on three dimensions where there is a classical scale symmetry. This is believed to be the worldvolume theory describing D2-branes within type IIA non-relativistic string theory and arises as a non-relativistic limit of maximally supersymmetric (relativistic) Yang-Mills. Unlike the relativistic case, we find that SGYM only admits perturbative dynamical excitations on the Coulomb branch. We show that in three dimensions the classical scale symmetry is broken at one-loop due to logarithmic divergences. These cannot be regulated without introducing additional marginal deformations, and we present evidence that the theory is asymptotically free once these are included. In particular our analysis suggests that three-dimensional SGYM flows to a strongly coupled non-relativistic M2-brane theory.

hep-th↗
arXiv2026

Evaluating Intellectual Property Guardrails of Generative Image Models: A Technical Report

Generative image models are capable of producing images that bear a strong resemblance to, or replicate, recognizable intellectual property (IP). In this technical report, we present a benchmark and automated evaluation pipeline to test for evidence of IP guardrails in generative image models along with the propensity for these models to generate images with recognizable IP. The IP categories we tested include fictional characters, celebrity likeness, and commercial logos and do not encompass the full range of IP which may be implicated by image generation models. We evaluated fourteen widely used text-to-image models, including three self-hosted open weights models and eleven private models. While all of the private models were observed to refuse generations at some level due to IP guardrails, the frequency of generation refusals varied substantially among models. The refusal rates also varied considerably across the different IP categories tested. Commercial logos were refused least frequently and were successfully generated at the highest rate, on average. Though the rate varies, all models tested readily generated images containing recognizable IP as of March 2026.

cs.CV↗
arXiv2026

Non-relativistic limits of $\mathcal N=4$ supersymmetric Yang-Mills theory and S-duality

We investigate non-relativistic limits of four-dimensional maximally supersymmetric Yang-Mills theory (4d MSYM) and their relation to the nonperturbative $\operatorname{SL}(2;\mathbb Z)$ S-duality of the relativistic theory. We construct a general family of non-relativistic limits using a Type IIB brane set-up with a D3-brane and $(p,q)$-strings and show that the resulting theories are topological deformations of supersymmetric Galilean Yang-Mills theory or quantum mechanics on the moduli space of BPS monopoles. The deformations of the Galilean Yang-Mills theory are the familiar $\theta$-term and a coupling to the monopole charge, while in the moduli space theory the only deformation is a $\theta$-term. This family of theories fit together into a three-dimensional moduli space with nontrivial topology, on which $\operatorname{PSL}(2;\mathbb Z)$-valued dualities act in a richer and more complex way than in the relativistic parent theory. In the Abelian case, we establish the duality directly using the path integral, while in the non-Abelian case we support our claim by matching the one-particle spectrum as well as the Galilean spacetime symmetries and electric/magnetic invertible one-form symmetries.

hep-th↗
arXiv2025

Forward and Backward Electron Acceleration by Radially Polarized Ultra-Intense Laser Focus Seeded By Field Ionization of High Charge States of Neon

Thanks to the fabrication of large aperture phase optics, ultra-intense relativistic laser plasma interaction (RLPI) experiments with complex polarization states are becoming feasible. In this work, we perform a computational investigation of direct acceleration of electrons produced during ionization of underdense neon gas using a tightly focused and radially polarized Petawatt-class short pulse lasers by numerically solving the relativistically invariant Lorentz equations, incorporating semi-classical tunneling ionization and Monte Carlo type sampling of the focal volume. The accelerated electrons energy gain increases at longer laser wavelengths and GeV energies are reached for electrons ionized from the neon inner shells, which are field ionized near the peak of the pulse. Backward acceleration of electrons is observed for a range of initial positions and phases of ionization of neon charge states. This apparent counterintuitive phenomenon is directly linked to the radial polarization state of the incident laser beam that results in a strong longitudinal electric field Ez when tightly focused, where electrons ionized near the focal center at the phase when Ez is pointed toward the forward propagation direction experiences an initial push in the backward direction. A parametric study of the phenomenon by varying laser parameters is presented, and a 3D particle in cell (PIC) simulation is considered to confirm the existence of this phenomenon.

physics.plasm-ph↗
arXiv2025

The M5-Brane Limit of Eleven-Dimensional Supergravity

We construct the M5-brane limit of eleven-dimensional supergravity. The resulting action is invariant under Galilean boosts and has a local scale symmetry. We also consider the limit of the equations of motion where we recover a Poisson-like equation arising from an M5-brane source but which does not follow from the non-relativistic action. We argue that the resulting theory describes gravitational fluctuations around a stack of M5-branes, represented by a trivial Minkowskian spacetime, but where the number of M5-branes is determined by the flux of a Lagrange multiplier field.

hep-th↗
arXiv2024

Reciprocal Non-Relativistic Decoupling Limits of String Theory and M-Theory

It has recently been realised that there are supersymmetry-preserving non-relativistic decoupling limits associated with each half-BPS object in String Theory and M-Theory. We argue that, given a $p$-brane and a $q$-brane for which there is a quarter-BPS intersecting configuration, the $p$-brane decoupling limit of the $q$-brane's worldvolume QFT is necessarily equal to the reciprocal $q$-brane decoupling limit of the $p$-brane's worldvolume theory. This is explicitly shown for the cases of the D0-D4 and D1-D3 systems. As an application of this idea, we use the reciprocal pair of limits in the M2-M5 system to describe the dynamics of interacting self-dual strings in the six-dimensional $\mathcal{N}=(2,0)$ theory. We also discuss the limits of the dual gravitational theories and, using symmetries, argue that the duality is maintained by the limits.

hep-th↗
arXiv2024

An Augmentation-based Model Re-adaptation Framework for Robust Image Segmentation

Image segmentation is a crucial task in computer vision, with wide-ranging applications in industry. The Segment Anything Model (SAM) has recently attracted intensive attention; however, its application in industrial inspection, particularly for segmenting commercial anti-counterfeit codes, remains challenging. Unlike open-source datasets, industrial settings often face issues such as small sample sizes and complex textures. Additionally, computational cost is a key concern due to the varying number of trainable parameters. To address these challenges, we propose an Augmentation-based Model Re-adaptation Framework (AMRF). This framework leverages data augmentation techniques during training to enhance the generalisation of segmentation models, allowing them to adapt to newly released datasets with temporal disparity. By observing segmentation masks from conventional models (FCN and U-Net) and a pre-trained SAM model, we determine a minimal augmentation set that optimally balances training efficiency and model performance. Our results demonstrate that the fine-tuned FCN surpasses its baseline by 3.29% and 3.02% in cropping accuracy, and 5.27% and 4.04% in classification accuracy on two temporally continuous datasets. Similarly, the fine-tuned U-Net improves upon its baseline by 7.34% and 4.94% in cropping, and 8.02% and 5.52% in classification. Both models outperform the top-performing SAM models (ViT-Large and ViT-Base) by an average of 11.75% and 9.01% in cropping accuracy, and 2.93% and 4.83% in classification accuracy, respectively.

cs.CV↗
arXiv2024

Turbulence Strength $C_n^2$ Estimation from Video using Physics-based Deep Learning

Images captured from a long distance suffer from dynamic image distortion due to turbulent flow of air cells with random temperatures, and thus refractive indices. This phenomenon, known as image dancing, is commonly characterized by its refractive-index structure constant $C_n^2$ as a measure of the turbulence strength. For many applications such as atmospheric forecast model, long-range/astronomy imaging, and aviation safety, optical communication technology, $C_n^2$ estimation is critical for accurately sensing the turbulent environment. Previous methods for $C_n^2$ estimation include estimation from meteorological data (temperature, relative humidity, wind shear, etc.) for single-point measurements, two-ended pathlength measurements from optical scintillometer for path-averaged $C_n^2$, and more recently estimating $C_n^2$ from passive video cameras for low cost and hardware complexity. In this paper, we present a comparative analysis of classical image gradient methods for $C_n^2$ estimation and modern deep learning-based methods leveraging convolutional neural networks. To enable this, we collect a dataset of video capture along with reference scintillometer measurements for ground truth, and we release this unique dataset to the scientific community. We observe that deep learning methods can achieve higher accuracy when trained on similar data, but suffer from generalization errors to other, unseen imagery as compared to classical methods. To overcome this trade-off, we present a novel physics-based network architecture that combines learned convolutional layers with a differentiable image gradient method that maintains high accuracy while being generalizable across image datasets.

cs.CV↗
arXiv2024

Non-Relativistic Intersecting Branes, Newton-Cartan Geometry and AdS/CFT

We discuss non-relativistic variants of four-dimensional ${\cal N}$=4 super-Yang-Mills theory obtained from generalised Newton-Cartan geometric limits of D3-branes in ten-dimensional spacetime. We argue that the natural interpretation of these limits is that they correspond to non-relativistic D1-branes or D3-branes intersecting the original D3-branes. The resulting gauge theories have dynamics that reduce to quantum mechanics on monopole moduli space or two-dimensional sigma-models on Hitchin moduli space respectively. We show that these theories possess interesting infinite-dimensional symmetries and we discuss the dual $AdS$ geometries.

hep-th↗
arXiv2024

How Quality Affects Deep Neural Networks in Fine-Grained Image Classification

In this paper, we propose a No-Reference Image Quality Assessment (NRIQA) guided cut-off point selection (CPS) strategy to enhance the performance of a fine-grained classification system. Scores given by existing NRIQA methods on the same image may vary and not be as independent of natural image augmentations as expected, which weakens their connection and explainability to fine-grained image classification. Taking the three most commonly adopted image augmentation configurations -- cropping, rotating, and blurring -- as the entry point, we formulate a two-step mechanism for selecting the most discriminative subset from a given image dataset by considering both the confidence of model predictions and the density distribution of image qualities over several NRIQA methods. Concretely, the cut-off points yielded by those methods are aggregated via majority voting to inform the process of image subset selection. The efficacy and efficiency of such a mechanism have been confirmed by comparing the models being trained on high-quality images against a combination of high- and low-quality ones, with a range of 0.7% to 4.2% improvement on a commercial product dataset in terms of mean accuracy through four deep neural classifiers. The robustness of the mechanism has been proven by the observations that all the selected high-quality images can work jointly with 70% low-quality images with 1.3% of classification precision sacrificed when using ResNet34 in an ablation study.

cs.CV↗
arXiv2024

Robust and Explainable Fine-Grained Visual Classification with Transfer Learning: A Dual-Carriageway Framework

In the realm of practical fine-grained visual classification applications rooted in deep learning, a common scenario involves training a model using a pre-existing dataset. Subsequently, a new dataset becomes available, prompting the desire to make a pivotal decision for achieving enhanced and leveraged inference performance on both sides: Should one opt to train datasets from scratch or fine-tune the model trained on the initial dataset using the newly released dataset? The existing literature reveals a lack of methods to systematically determine the optimal training strategy, necessitating explainability. To this end, we present an automatic best-suit training solution searching framework, the Dual-Carriageway Framework (DCF), to fill this gap. DCF benefits from the design of a dual-direction search (starting from the pre-existing or the newly released dataset) where five different training settings are enforced. In addition, DCF is not only capable of figuring out the optimal training strategy with the capability of avoiding overfitting but also yields built-in quantitative and visual explanations derived from the actual input and weights of the trained model. We validated DCF's effectiveness through experiments with three convolutional neural networks (ResNet18, ResNet34 and Inception-v3) on two temporally continued commercial product datasets. Results showed fine-tuning pathways outperformed training-from-scratch ones by up to 2.13% and 1.23% on the pre-existing and new datasets, respectively, in terms of mean accuracy. Furthermore, DCF identified reflection padding as the superior padding method, enhancing testing accuracy by 3.72% on average. This framework stands out for its potential to guide the development of robust and explainable AI solutions in fine-grained visual classification tasks.

cs.CV↗
arXiv2024

Non-Relativistic M2-Branes and the AdS/CFT Correspondence

A non-relativistic limit of the AdS/CFT correspondence is studied in the context of M2-branes. On the field theory side this corresponds to a near-BPS limit of ABJM that localises onto solutions of Hitchin's equations. It is shown that the symmetries of the theory include an infinite-dimensional enhancement of the spatial symmetry algebra corresponding to time-dependent holomorphic transformations. Taking the limit of the gravitational dual splits the geometry into three 'large' directions and eight 'small' directions and corresponds to the Membrane-Newton-Cartan limit of eleven-dimensional supergravity. This has the effect of reducing the $AdS_4$ factor to an $AdS_2$ factor for the near-horizon limit of the M2-brane metric. Evidence is presented that the duality is maintained after the limit.

hep-th↗
arXiv2023

Fermions with $SU(1,n)$ Spacetime Symmetry

We construct theories of free fermions in $(2n-1)$-dimensions with $SU(1,n)$ spacetime symmetry from the null reduction of fermions on a $2n$-dimensional $\Omega$-deformed Minkowski background for $n=2$ and $n=3$. These play a role in the 5d $SU(1,3)$-invariant theories that are conjectured to offer a full description of certain 6d superconformal field theories. We find the $(2n-1)$-dimensional manifestation of the supersymmetry of a free $2n$-dimensional boson-fermion system, which we use to fix the fermion two-point functions. It is then shown that the full $2n$-dimensional two-point function can be recovered through resummation. Limits of the theories are considered, and it is observed that both Galilean and Carrollian field theories appear in different regimes. We confirm that the correlation functions obey the $SU(1,n)$ Ward identities and the representations of the fermions under this group are discussed.

hep-th↗
arXiv2022

RG Flows and Symmetry Enhancement in Five-Dimensional Lifshitz Gauge Theories

Lagrangian gauge theories with a z=2 Lifshitz scaling provide a family of interacting, asymptotically free five-dimensional field theories. We examine some of their quantum properties, extending previous results to include matter. We present no-go theorems that, in the absence of constraints, such theories cannot admit a spinorial supersymmetry or a boost symmetry. However, we argue that there exist renormalization group flows whose fixed points can admit supersymmetry and boosts, i.e. super-Schrodinger symmetry. We also present examples of Lifshitz gauge theories with a scalar supersymmetry.

hep-th↗