arXiv subjects

David A. Smith

Publications and source records attributed to David A. Smith.

At least 19 recordsLinked to original sources

arXiv2026

The GMRT High-Resolution Southern Sky Survey for pulsars and transients -- VIII: Orbital Variability and the Evolution of a 1-Day He-WD Millisecond Pulsar J2101-4802

We present timing and orbital phase-resolved polarimetry of the millisecond pulsar (MSP) J2101$-$4802, having a spin period of 9.48~ms and dispersion measure (DM) $25.05\ \mathrm{pc\ cm^{-3}}$ discovered with the Giant Meter Radio Telescope (GMRT). From the phase-connected timing of this MSP spanning 3.7 years, we identify that PSR J2101-4802 is in a $\sim$1-day binary orbit with a likely helium-white-dwarf (He-WD) companion having a median companion mass of $\simeq0.15\, M_\odot$, consistent with canonical recycling in the Galactic field. The timing solution further reveals an unusually large orbital period derivative, $\dot{P}_b$ ($\sim10^{-11}\,{\rm s\,s}^{-1}$), compared to typical Galactic-field MSP--HeWD binaries, which cannot be explained by the contributions from kinematic effects (Shklovskii and Galactic acceleration) or general-relativistic damping. Using wideband, full-Stokes observations, we also trace the linear and circular polarization variation across the orbital phase and fit a rotating-vector model (RVM) to its position-angle swing across the pulse phase, yielding constraints on the emission geometry (magnetic inclination and impact angle) of this system. The combination of a $\sim$1-day orbit, $\sim0.15\,M_\odot$ companion, modest spin-down power, unusually large $\dot{P}_b$, and phase-locked magnetized intrabinary plasma signatures suggests that PSR~J2101$-$4802 represents a transitional system linking redback-like spiders to detached He--WD MSP binaries.

astro-ph.HE↗

arXiv2026

Randomized Masked Finetuning: An Efficient Way to Mitigate Memorization of PIIs in LLMs

The current literature on memorization in Natural Language Models, especially Large Language Models (LLMs), poses severe security and privacy risks, as models tend to memorize personally identifying information (PIIs) from training data. We introduce Randomized Masked Fine-Tuning (RMFT), a novel privacy-preserving fine-tuning technique that reduces PII memorization while minimizing performance impact. Using the Enron Email Dataset, we demonstrate that RMFT achieves an 80.81% reduction in Total Extraction Rate and 80.17% reduction in Seen Extraction Rate compared to baseline fine-tuning, outperforming deduplication methods while maintaining only a 5.73% increase in perplexity. We present MaxTER, a Pareto-optimal evaluation framework for assessing privacy-utility tradeoffs, and show the performance of RMFT vs Deduplication by Area Under The Response Curve (AURC) metric.

cs.CL↗

arXiv2026

Wireless Sensing of Temperature, Strain and Crack Growth in 3D-Printed Metal Structures via Magneto-Responsive Inclusions

This study demonstrates the first realization of wireless strain, temperature and crack growth sensing within 3D-printed metallic structures using standard electromagnetic inspection hardware. This establishes a path toward need-based maintenance for parts operating in harsh environments driven by accurate, real-time damage assessments instead of relying on regularly scheduled maintenance teardowns. To this end, we encapsulate and embed magnetoelastic and thermomagnetic materials during additive manufacturing. Mechanical and thermal stimuli affect the magnetic permeability of the embedded materials, which modulates the flux through a coil placed on or near the part's surface. We demonstrate strain sensing accurate to +/-27x10-6 and temperature sensing accurate to +/-0.75 oC. We highlight these sensors' capabilities by detecting the onset of plasticity and fatigue-driven crack growth thousands of cycles before critical failure.

cs.ET↗

arXiv2026

Memorization Dynamics in Knowledge Distillation for Language Models

Knowledge Distillation (KD) is increasingly adopted to transfer capabilities from large language models to smaller ones, offering significant improvements in efficiency and utility while often surpassing standard fine-tuning. Beyond performance, KD is also explored as a privacy-preserving mechanism to mitigate the risk of training data leakage. While training data memorization has been extensively studied in standard pre-training and fine-tuning settings, its dynamics in a knowledge distillation setup remain poorly understood. In this work, we study memorization across the KD pipeline using three large language model (LLM) families (Pythia, OLMo-2, Qwen-3) and three datasets (FineWeb, Wikitext, Nemotron-CC-v2). We find: (1) distilled models memorize significantly less training data than standard fine-tuning (reducing memorization by more than 50%); (2) some examples are inherently easier to memorize and account for a large fraction of memorization during distillation (over ~95%); (3) student memorization is predictable prior to distillation using features based on zlib entropy, KL divergence, and perplexity; and (4) while soft and hard distillation have similar overall memorization rates, hard distillation poses a greater risk: it inherits $2.7\times$ more teacher-specific examples than soft distillation. Overall, we demonstrate that distillation can provide both improved generalization and reduced memorization risks compared to standard fine-tuning.

cs.CL↗

arXiv2025

Privacy Ripple Effects from Adding or Removing Personal Information in Language Model Training

Due to the sensitive nature of personally identifiable information (PII), its owners may have the authority to control its inclusion or request its removal from large-language model (LLM) training. Beyond this, PII may be added or removed from training datasets due to evolving dataset curation techniques, because they were newly scraped for retraining, or because they were included in a new downstream fine-tuning stage. We find that the amount and ease of PII memorization is a dynamic property of a model that evolves throughout training pipelines and depends on commonly altered design choices. We characterize three such novel phenomena: (1) similar-appearing PII seen later in training can elicit memorization of earlier-seen sequences in what we call assisted memorization, and this is a significant factor (in our settings, up to 1/3); (2) adding PII can increase memorization of other PII significantly (in our settings, as much as $\approx\!7.5\times$); and (3) removing PII can lead to other PII being memorized. Model creators should consider these first- and second-order privacy risks when training models to avoid the risk of new PII regurgitation.

cs.CL↗

arXiv2024

Learning from Many Voices: Literary MT Using Multi-Reference Human and Synthetic Data

Multiple valid translations of a single literary work naturally exist. We investigate strategies for leveraging these multi-reference datasets to improve literary machine translation. We propose a filtering framework based on semantic similarity to identify source texts whose references display meaningful variation while remaining faithful. We find that fine-tuning with medium to high semantic similarity data substantially outperforms low semantic similarity data. Moreover, using medium and high semantic similarity data achieves comparable or better performance than using the full unfiltered data. Synthetic translations generated by LLMs are economical and convenient alternatives to human expert translations; however, we find fine-tuning on human expert translations outperforms fine-tuning on synthetically augmented data in automatic metrics and human evaluations, demonstrating the indispensable value of human expert translations for fine-tuning literary machine translation models.

cs.CL↗

arXiv2024

Vertically Graded Fe-Ni Alloys with Low Damping and a Sizeable Spin-Orbit Torque

Energy-efficient spintronic devices require a large spin-orbit torque (SOT) and low damping to excite magnetic precession. In conventional devices with heavy-metal/ferromagnet bilayers, reducing the ferromagnet thickness to $\sim$1 nm enhances the SOT but dramatically increases damping. Here, we investigate an alternative approach based on a 10 nm thick single-layer ferromagnet to attain both low damping and a sizable SOT. Instead of relying on a single interface, we continuously break the bulk inversion symmetry with a vertical compositional gradient of two ferromagnetic elements: Fe with low intrinsic damping and Ni with sizable spin-orbit coupling. We find low effective damping parameters of $α_\mathrm{eff} < 5\times10^{-3}$ in the FeNi alloy films, despite the steep compositional gradients. Moreover, we reveal a sizable anti-damping SOT efficiency of $|θ_\mathrm{DL}| \approx 0.05$, even without an intentional compositional gradient. Through depth-resolved x-ray diffraction, we identify a lattice strain gradient as crucial symmetry breaking that underpins the SOT. Our findings provide fresh insights into damping and SOTs in single-layer ferromagnets for power-efficient spintronic devices.

cond-mat.mtrl-sci↗

arXiv2024

Large-Amplitude, Easy-Plane Spin-Orbit Torque Oscillators Driven by Out-of-Plane Spin Current: A Micromagnetic Study

Spin torque oscillators are spintronic devices that generate a periodic output signal from a non-periodic input, making them promising candidates for applications like microwave communications and neuromorphic computing. However, traditional spin torque oscillators suffer from a limited precessional cone angle and thermal stability, as well as a need for an applied bias magnetic field. We use micromagnetic simulations to demonstrate a novel spin torque oscillator that relies on spin-orbit effects in ferromagnets to overcome these limitations. The key mechanism behind this oscillator is the generation of an out-of-plane spin current, in which both the spin flow and the spin orientation are out-of-plane. The torque from this spin current enables easy-plane coherent magnetic precession with a large cone angle and high thermal stability over a micron-scale lateral area. Moreover, the precession occurs about an internal field in the free layer, thereby eliminating the need for an external bias field. We demonstrate the feasibility of an easy-plane spin-orbit torque oscillator at room temperature over a wide parameter space, including the ratio of the out-of-plane spin current to the conventional spin-Hall spin current, presenting exciting possibilities for this novel spintronic device.

cond-mat.mes-hall↗

arXiv2024

Mind the Gap: Analyzing Lacunae with Transformer-Based Transcription

Historical documents frequently suffer from damage and inconsistencies, including missing or illegible text resulting from issues such as holes, ink problems, and storage damage. These missing portions or gaps are referred to as lacunae. In this study, we employ transformer-based optical character recognition (OCR) models trained on synthetic data containing lacunae in a supervised manner. We demonstrate their effectiveness in detecting and restoring lacunae, achieving a success rate of 65%, compared to a base model lacking knowledge of lacunae, which achieves only 5% restoration. Additionally, we investigate the mechanistic properties of the model, such as the log probability of transcription, which can identify lacunae and other errors (e.g., mistranscriptions due to complex writing or ink issues) in line images without directly inspecting the image. This capability could be valuable for scholars seeking to distinguish images containing lacunae or errors from clean ones. Although we explore the potential of attention mechanisms in flagging lacunae and transcription errors, our findings suggest it is not a significant factor. Our work highlights a promising direction in utilizing transformer-based OCR models for restoring or analyzing damaged historical documents.

cs.CV↗

arXiv2024

Revivals, or the Talbot effect, for the Airy equation

We study Dirichlet-type problems for the simplest third-order linear dispersive PDE, often referred to as the Airy equation. Such problems have not been extensively studied, perhaps due to the complexity of the spectral structure of the spatial operator. Our specific interest is to determine whether the peculiar phenomenon of revivals, also known as Talbot effect, is supported by these boundary conditions, which for third order problems are not reducible to periodic ones. We prove that this is the case only for a very special choice of the boundary conditions, for which a new type of weak cusp revival phenomenon has been recently discovered. We also give some new results on the functional class of the solution for other cases.

math.AP↗

arXiv2024

Jumps and cusps: a new revival effect in local dispersive PDEs

We study the presence of a non-trivial revival effect in the solution of linear dispersive boundary value problems for two benchmark models which arise in applications: the Airy equation and the dislocated Laplacian Schr{ö}dinger equation. In both cases, we consider boundary conditions of Dirichlet-type. We prove that, at suitable times, jump discontinuities in the initial profile are revived in the solution not only as jump discontinuities but also as logarithmic cusp singularities. We explicitly describe these singularities and show that their formation is due to interactions between the symmetries of the underlying spatial operators with the periodic Hilbert transform.

math.AP↗

arXiv2023

The Third Fermi Large Area Telescope Catalog of Gamma-ray Pulsars

We present 294 pulsars found in GeV data from the Large Area Telescope (LAT) on the Fermi Gamma-ray Space Telescope. Another 33 millisecond pulsars (MSPs) discovered in deep radio searches of LAT sources will likely reveal pulsations once phase-connected rotation ephemerides are achieved. A further dozen optical and/or X-ray binary systems co-located with LAT sources also likely harbor gamma-ray MSPs. This catalog thus reports roughly 340 gamma-ray pulsars and candidates, 10% of all known pulsars, compared to $\leq 11$ known before Fermi. Half of the gamma-ray pulsars are young. Of these, the half that are undetected in radio have a broader Galactic latitude distribution than the young radio-loud pulsars. The others are MSPs, with 6 undetected in radio. Overall, >235 are bright enough above 50 MeV to fit the pulse profile, the energy spectrum, or both. For the common two-peaked profiles, the gamma-ray peak closest to the magnetic pole crossing generally has a softer spectrum. The spectral energy distributions tend to narrow as the spindown power $\dot E$ decreases to its observed minimum near $10^{33}$ erg s$^{-1}$, approaching the shape for synchrotron radiation from monoenergetic electrons. We calculate gamma-ray luminosities when distances are available. Our all-sky gamma-ray sensitivity map is useful for population syntheses. The electronic catalog version provides gamma-ray pulsar ephemerides, properties and fit results to guide and be compared with modeling results.

astro-ph.HE↗

arXiv2023

Suppression of Spin Pumping at Metal Interfaces

An electrically conductive metal typically transmits or absorbs a spin current. Here, we report on evidence that interfacing two metal thin films can suppress spin transmission and absorption. We examine spin pumping in ferromagnet/spacer/ferromagnet heterostructures, in which the spacer -- consisting of metallic Cu and Cr thin films -- separates the ferromagnetic spin-source and spin-sink layers. The Cu/Cr spacer largely suppresses spin pumping -- i.e., neither transmitting nor absorbing a significant amount of spin current -- even though Cu or Cr alone transmits a sizable spin current. The antiferromagnetism of Cr is not essential for the suppression of spin pumping, as we observe similar suppression with Cu/V spacers where V is a nonmagnetic analogue of Cr. We speculate that diverse combinations of spin-transparent metals may form interfaces that suppress spin pumping, although the underlying mechanism remains unclear. Our work may stimulate a new perspective on understanding and engineering spin transport in metallic multilayers.

cond-mat.mtrl-sci↗

arXiv2023

Composition and Deformance: Measuring Imageability with a Text-to-Image Model

Although psycholinguists and psychologists have long studied the tendency of linguistic strings to evoke mental images in hearers or readers, most computational studies have applied this concept of imageability only to isolated words. Using recent developments in text-to-image generation models, such as DALLE mini, we propose computational methods that use generated images to measure the imageability of both single English words and connected text. We sample text prompts for image generation from three corpora: human-generated image captions, news article sentences, and poem lines. We subject these prompts to different deformances to examine the model's ability to detect changes in imageability caused by compositional change. We find high correlation between the proposed computational measures of imageability and human judgments of individual words. We also find the proposed measures more consistently respond to changes in compositionality than baseline approaches. We discuss possible effects of model training and implications for the study of compositionality in text-to-image models.

cs.CL↗

arXiv2023

Adapting Transformer Language Models for Predictive Typing in Brain-Computer Interfaces

Brain-computer interfaces (BCI) are an important mode of alternative and augmentative communication for many people. Unlike keyboards, many BCI systems do not display even the 26 letters of English at one time, let alone all the symbols in more complex systems. Using language models to make character-level predictions, therefore, can greatly speed up BCI typing (Ghosh and Kristensson, 2017). While most existing BCI systems employ character n-gram models or no LM at all, this paper adapts several wordpiece-level Transformer LMs to make character predictions and evaluates them on typing tasks. GPT-2 fares best on clean text, but different LMs react differently to noisy histories. We further analyze the effect of character positions in a word and context lengths.

cs.CL↗

arXiv2022

The SPAN512 mid-latitude pulsar survey at the Nançay Radio Telescope

The large number of ongoing surveys for pulsars and transients at various radio observatories is motivated by the science obtained from these sources. Timing and polarisation analysis of relativistic binaries can place strong constraints on theories of gravity. The observation of a growing number of millisecond pulsars (MSPs) spread over the celestial sphere may allow the detection of a stochastic gravitational wave background arising from supermassive black hole binaries. A more complete sample of young pulsars improves our knowledge of neutron star birth and evolution. Transients such as fast radio bursts can serve to probe the intergalactic medium. The SPAN512 pulsar survey covers intermediate Galactic latitudes using the L-band receiver of the Nançay Radio Telescope (NRT). The survey covers 224 sq. deg. of the sky for a total exposure time of 2200 h. Population syntheses predict the discovery of 3 to 19 new normal pulsars and a few MSPs. We present detailed modelling of the NRT beam with its L-band receiver and its sensitivity which we used to precisely assess the expected survey yield. We used the flexible Pulsar Arecibo L-band Feed Array data processing pipeline to search the 47 TB of SPAN512 data for pulsars and transients. The SPAN512 survey discovered two new MSPs and one new middle-aged pulsar. We focus on the analysis of the 2.4-ms spin period pulsar J2205+6012 for which we also report the detection of gamma-ray pulsations. Its narrow pulse width (35 $μ$s at an observing frequency of 2.55 GHz) allows for sub-microsecond timing precision over 8 years, with exciting prospects for pulsar timing array programs.

astro-ph.HE↗

arXiv2022

Absorption of Transverse Spin Current in Ferromagnetic NiCu: Dominance of Bulk Dephasing over Spin-Flip Scattering

In ferromagnetic metals, transverse spin currents are thought to be absorbed via dephasing -- i.e., destructive interference of spins precessing about the strong exchange field. Yet, due to the ultrashort coherence length of $\approx$1 nm in typical ferromagnetic thin films, it is difficult to distinguish dephasing in the bulk from spin-flip scattering at the interface. Here, to assess which mechanism dominates, we examine transverse spin-current absorption in ferromagnetic NiCu alloy films with reduced exchange fields. We observe that the coherence length increases with decreasing Curie temperature, as weaker dephasing in the film bulk slows down spin absorption. Moreover, nonmagnetic Cu impurities do not diminish the efficiency of spin-transfer torque from the absorbed spin current. Our findings affirm that transverse spin current is predominantly absorbed by dephasing inside the nanometer-thick ferromagnetic metals, even with high impurity contents.

cond-mat.mes-hall↗

arXiv2022

Room-Temperature Intrinsic and Extrinsic Damping in Polycrystalline Fe Thin Films

We examine room-temperature magnetic relaxation in polycrystalline Fe films. Out-of-plane ferromagnetic resonance (FMR) measurements reveal Gilbert damping parameters of $\approx$ 0.0024 for Fe films with thicknesses of 4-25 nm, regardless of their microstructural properties. The remarkable invariance with film microstructure strongly suggests that intrinsic Gilbert damping in polycrystalline metals at room temperature is a local property of nanoscale crystal grains, with limited impact from grain boundaries and film roughness. By contrast, the in-plane FMR linewidths of the Fe films exhibit distinct nonlinear frequency dependences, indicating the presence of strong extrinsic damping. To fit our in-plane FMR data, we have used a grain-to-grain two-magnon scattering model with two types of correlation functions aimed at describing the spatial distribution of inhomogeneities in the film. However, neither of the two correlation functions is able to reproduce the experimental data quantitatively with physically reasonable parameters. Our findings advance the fundamental understanding of intrinsic Gilbert damping in structurally disordered films, while demonstrating the need for a deeper examination of how microstructural disorder governs extrinsic damping.

cond-mat.mtrl-sci↗