Search

This paper presents a review of research on the effectiveness of educational television compared to traditional face-to-face instruction. The studies reviewed are presented under seven rubrics: TV as a catalyst for learning; two-way TV; use of commercial TV shows; simulation of a real situation; TV integrated as part of the classroom lecture; TV courses instead of lecture in the classroom; and TV courses in the home. The author concludes that while TV has proved effective for teaching basic knowledge, it is deficient for teaching cognitive skills requiring more than "level I" knowledge. A bibliography is appended (the latest reference is to a 1975 publication). (BB)

Robust medical Machine Learning (ML) models have the potential to revolutionize healthcare by accelerating clinical research, improving workflows and outcomes, and producing novel insights or capabilities. Developing such ML models from scratch is cost prohibitive and requires substantial compute, data, and time (e.g., expert labeling). To address these challenges, we introduce Health AI Developer Foundations (HAI-DEF), a suite of pre-trained, domain-specific foundation models, tools, and recipes to accelerate building ML for health applications. The models cover various modalities and domains, including radiology (X-rays and computed tomography), histopathology, dermatological imaging, and audio. These models provide domain specific embeddings that facilitate AI development with less labeled data, shorter training times, and reduced computational costs compared to traditional approaches. In addition, we utilize a common interface and style across these models, and prioritize usability to enable developers to integrate HAI-DEF efficiently. We present model evaluations across various tasks and conclude with a discussion of their application and evaluation, covering the importance of ensuring efficacy, fairness, and equity. Finally, while HAI-DEF and specifically the foundation models lower the barrier to entry for ML in healthcare, we emphasize the importance of validation with problem- and population-specific data for each desired usage setting. This technical report will be updated over time as more modalities and features are added., Comment: 16 pages, 8 figures

Adiabatic quantum-flux-parametron (AQFP) logic is a proven energy-efficient superconductor technology for various applications. To address the scalability challenges, we investigated AQFP shift registers with the AQFP footprint area reduced by 25% with respect to prior work and with more than 2x denser overall designs obtained by eliminating the previously used free space between the AQFPs. We also investigated AQFP cells with different designs of flux trapping moats in the superconducting ground plane as well as compact AQFP cells that took advantage of the smaller feature sizes available in the new fabrication process, SFQ5ee+, at MIT Lincoln Laboratory. This new process features nine planarized Nb layers with a 0.25 $\mu$m minimum linewidth. The fabricated circuits were tested in a liquid He probe and in a closed-cycle cryocooler using a controlled cooling rate through the superconducting critical temperature. Using multiple thermal cycles, we investigated flux trapping in the dense AQFP shift registers as well as in the registers using the old (sparse) AQFP designs at two levels of the residual magnetic field, about 0.53 $\mu$T and about 1.2 $\mu$T. The sparse designs demonstrated 95% to almost 100% probability of operation after the cooldown and very wide operation margins, although the flux trapping probability was increasing with circuit complexities. The margins were similarly wide in the newer dense designs, but flux trapping probability that rendered the registers nonoperational was significantly, by an order of magnitude, higher in the denser circuits and was also very sensitive to the moats' shape and location. Our findings indicate that AQFP circuits are amendable to increasing the scale of integration and further densification, but a careful moat design and optimization are required to reduce flux trapping effects in the dense AQFP circuits., Comment: 5 pages, 10 figures, 2 tables, 18 references. Presented at Applied Superconductivity Conference, ASC2024, 1-6 September 2024, Salt Lake City, UT, USA

If the probability model is correctly specified, then we can estimate the covariance matrix of the asymptotic maximum likelihood estimate distribution using either the first or second derivatives of the likelihood function. Therefore, if the determinants of these two different covariance matrix estimation formulas differ this indicates model misspecification. This misspecification detection strategy is the basis of the Determinant Information Matrix Test ($GIMT_{Det}$). To investigate the performance of the $GIMT_{Det}$, a Deterministic Input Noisy And gate (DINA) Cognitive Diagnostic Model (CDM) was fit to the Fraction-Subtraction dataset. Next, various misspecified versions of the original DINA CDM were fit to bootstrap data sets generated by sampling from the original fitted DINA CDM. The $GIMT_{Det}$ showed good discrimination performance for larger levels of misspecification. In addition, the $GIMT_{Det}$ did not detect model misspecification when model misspecification was not present and additionally did not detect model misspecification when the level of misspecification was very low. However, the $GIMT_{Det}$ discrimation performance was highly variable across different misspecification strategies when the misspecification level was moderately sized. The proposed new misspecification detection methodology is promising but additional empirical studies are required to further characterize its strengths and limitations.

The number density and redshift evolution of optically selected galaxy clusters offer an independent measurement of the amplitude of matter fluctuations, $S_8$. However, recent results have shown that clusters chosen by the redMaPPer algorithm show richness-dependent biases that affect the weak lensing signals and number densities of clusters, increasing uncertainty in the cluster mass calibration and reducing their constraining power. In this work, we evaluate an alternative cluster proxy, outskirt stellar mass, $M_{\textrm{out}}$, defined as the total stellar mass within a $[50,100]$ kpc envelope centered on a massive galaxy. This proxy exhibits scatter comparable to redMaPPer richness, $\lambda$, but is less likely to be subject to projection effects. We compare the Dark Energy Survey Year 3 redMaPPer cluster catalog with a $M_{\textrm{out}}$ selected cluster sample from the Hyper-Suprime Camera survey. We use weak lensing measurements to quantify and compare the scatter of $M_{\textrm{out}}$ and $\lambda$ with halo mass. Our results show $M_{\textrm{out}}$ has a scatter consistent with $\lambda$, with a similar halo mass dependence, and that both proxies contain unique information about the underlying halo mass. We find $\lambda$-selected samples introduce features into the measured $\Delta \Sigma$ signal that are not well fit by a log-normal scatter only model, absent in $M_{\textrm{out}}$ selected samples. Our findings suggest that $M_{\textrm{out}}$ offers an alternative for cluster selection with more easily calibrated selection biases, at least at the generally lower richnesses probed here. Combining both proxies may yield a mass proxy with a lower scatter and more tractable selection biases, enabling the use of lower mass clusters in cosmology. Finally, we find the scatter and slope in the $\lambda-M_{\textrm{out}}$ scaling relation to be $0.49 \pm 0.02$ and $0.38 \pm 0.09$., Comment: 22 pages, 8 figures, 4 tables, submitted to PRD

As our understanding of autism and ableism continues to increase, so does our understanding of ableist language towards autistic people. Such language poses a significant challenge in NLP research due to its subtle and context-dependent nature. Yet, detecting anti-autistic ableist language remains underexplored, with existing NLP tools often failing to capture its nuanced expressions. We present AUTALIC, the first benchmark dataset dedicated to the detection of anti-autistic ableist language in context, addressing a significant gap in the field. The dataset comprises 2,400 autism-related sentences collected from Reddit, accompanied by surrounding context, and is annotated by trained experts with backgrounds in neurodiversity. Our comprehensive evaluation reveals that current language models, including state-of-the-art LLMs, struggle to reliably identify anti-autistic ableism and align with human judgments, underscoring their limitations in this domain. We publicly release AUTALIC along with the individual annotations which serve as a valuable resource to researchers working on ableism, neurodiversity, and also studying disagreements in annotation tasks. This dataset serves as a crucial step towards developing more inclusive and context-aware NLP systems that better reflect diverse perspectives., Comment: 9 pages, 5 figures, 7 tables

The Charting Cluster Construction with VUDS and ORELSE (C3VO) survey is an ongoing imaging and spectroscopic campaign aiming to map out the growth of structure up to $z\sim5$ and was born from the combination of the VIMOS Ultra Deep Survey (VUDS) and the Observations of Redshift Evolution in Large-Scale Environments (ORELSE) survey. As we previously accomplished with the ORELSE survey, we apply our technique known as Voronoi tessellation Monte-Carlo (VMC) mapping to search for serendipitous galaxy overdensities at $2 14$. In the C3VO fields, we were able to recover many of the previously known structures in the literature as well as find hundreds of new overdensity candidates, once again demonstrating the powerful capabilities of VMC mapping when applied to wide-field optical and infrared galaxy evolution surveys at ever higher redshifts., Comment: 36 pages, 16 figures, 5 tables, submitted to ApJ

High-precision pulsar timing is highly dependent on precise and accurate modeling of any effects that impact the data. It was shown that commonly used Solar Wind models do not accurately account for variability in the amplitude of the Solar wind on both short and long time scales. In this study, we test and validate a new, cutting-edge Solar wind modeling method included in the \texttt{enterprise} software suite through extended simulations, and we apply it to investigate temporal variability in LOFAR data. Our model testing scheme in itself provides an invaluable asset for pulsar timing array (PTA) experiments. As improperly accounting for the solar wind signature in pulsar data can induce false-positive signals, it is of fundamental importance to include in any such investigations. We employ a Bayesian approach utilizing a continuously varying Gaussian process to model the solar wind referred to as Solar Wind Gaussian Process (SWGP). We conduct noise analysis on eight pulsars from the LOFAR dataset with most pulsars having a timespan of $\sim 11$ years encompassing one full solar activity cycle. Our analysis reveals a strong correlation between the electron density at 1 AU and the ecliptic latitude (ELAT) of the pulsar. Pulsars with $|ELAT|< 3^{\circ}$ exhibit significantly higher average electron densities. We observe distinct temporal patterns in electron densities in different pulsars. In particular, pulsars within $|ELAT|< 3^{\circ}$ exhibit similar temporal variations, while the electron densities of those outside this range correlate with the solar activity cycle. The continuous variability in electron density offered in this model represents a substantial improvement over previous models, which assume a single value for piece-wise bins of time. This advancement holds promise for solar wind modeling in future International Pulsar Timing Array data combinations., Comment: Accepted in Section 9. Sun and the Heliosphere, Astronomy and Astrophysics

We conduct a study on the relationship between galaxy environments and their active galactic nuclei (AGN) activity at high redshifts ($2.010^{12.8}M_\odot$) overdense ($\sigma_\delta>5.0$) peak, normalized with respect to the size of the peak. These findings indicate that the prevalence of AGN activity is highly dependent on the environment in which a host galaxy resides, even at early times in the formation history of the Universe., Comment: 15 pages, 11 figures, submitted to A&A, comments welcome

Using a sample of 2800 galaxy clusters identified in the Dark Energy Survey across the redshift range $0.20 < z < 0.60$, we characterize the hierarchical assembly of Bright Central Galaxies (BCGs) and the surrounding intracluster light (ICL). To quantify hierarchical formation we use the stellar mass - halo mass (SMHM) relation for the BCG+ICL system and incorporate the magnitude gap (M14), the difference in brightness between the BCG (measured within 30kpc) and 4th brightest cluster member galaxy within 0.5 $R_{200,c}$. The inclusion of M14, which traces BCG hierarchical growth, increases the slope and decreases the intrinsic scatter in the SMHM relation, highlighting that it is a latent variable within the BCG+ICL SMHM relation. Moreover, the correlation with M14 decreases at large radii from the BCG's centre. However, the stellar light within the BCG+ICL transition region (30kpc - 80kpc) most strongly correlates with the dark matter halo mass and has a statistically significant correlation with M14. As the light in the transition region and M14 are independent measurements, the transition region may grow as a result of the BCG's hierarchical two-phase formation. Additionally, as M14 and ICL result from hierarchical growth, we use a stacked sample and find that clusters with large M14 values are characterized by larger ICL and BCG+ICL fractions, which illustrates that the merger processes that build the BCG stellar mass also grow the ICL. Furthermore, this may suggest that M14 combined with the ICL fraction can be used as a method to identify dynamically relaxed clusters., Comment: 16 pages, 5 Figures, submitted to MNRAS on 8/30/2024

Microscopic interpretation of histopathology images underlies many important diagnostic and treatment decisions. While advances in vision-language modeling raise new opportunities for analysis of such images, the gigapixel-scale size of whole slide images (WSIs) introduces unique challenges. Additionally, pathology reports simultaneously highlight key findings from small regions while also aggregating interpretation across multiple slides, often making it difficult to create robust image-text pairs. As such, pathology reports remain a largely untapped source of supervision in computational pathology, with most efforts relying on region-of-interest annotations or self-supervision at the patch-level. In this work, we develop a vision-language model based on the BLIP-2 framework using WSIs paired with curated text from pathology reports. This enables applications utilizing a shared image-text embedding space, such as text or image retrieval for finding cases of interest, as well as integration of the WSI encoder with a frozen large language model (LLM) for WSI-based generative text capabilities such as report generation or AI-in-the-loop interactions. We utilize a de-identified dataset of over 350,000 WSIs and diagnostic text pairs, spanning a wide range of diagnoses, procedure types, and tissue types. We present pathologist evaluation of text generation and text retrieval using WSI embeddings, as well as results for WSI classification and workflow prioritization (slide-level triaging). Model-generated text for WSIs was rated by pathologists as accurate, without clinically significant error or omission, for 78% of WSIs on average. This work demonstrates exciting potential capabilities for language-aligned WSI embeddings., Comment: 9 main pages and 19 pages of supplemental material; 3 main tables, 3 main figures and 11 supplemental tables, 7 supplemental figures

We present an analysis of the number density of galaxies as a function of stellar mass (i.e., the stellar mass function, SMF) in the COSMOS field at z~3.3, making a comparison between the SMF in overdense environments and the SMF in the coeval field. In particular, this region contains the Elent\'ari proto-supercluster, a system of 6 extended overdensities spanning ~70 cMpc on a side. A clear difference is seen in the high-mass slope of these SMFs, with overdense regions showing an increase in the ratio of high-mass galaxies to low-mass galaxies relative to the field, indicating a more rapid build-up of stellar mass in overdense environments. This result qualitatively agrees with analyses of clusters at z~1, though the differences between protocluster and field SMFs at z~3.3 are smaller. While this is consistent with overdensities enhancing the evolution of their member galaxies, potentially through increased merger rates, whether this enhancement begins in protocluster environments or even earlier in group environments is still unclear. Though the measured fractions of quiescent galaxies between the field and overdense environments do not vary significantly, implying that this stellar mass enhancement is ongoing and any starbursts triggered by merger activity have not yet quenched, we note that spectroscopic observations are biased towards star-forming populations, particularly for low-mass galaxies. If mergers are indeed responsible, high resolution imaging of Elent\'ari and similar structures at these early epochs should then reveal increased merger rates relative to the field. Larger samples of well-characterized overdensities are necessary to draw broader conclusions in these areas., Comment: 15 pages, 8 figures, plus references and appendices. Submitted to ApJ, posted after responding to referee comments. Further comments welcome

We present the first analysis of the Euclid Early Release Observations (ERO) program that targets fields around two lensing clusters, Abell 2390 and Abell 2764. We use VIS and NISP imaging to produce photometric catalogs for a total of $\sim 500\,000$ objects. The imaging data reach a $5\,\sigma$ typical depth in the range 25.1-25.4 AB in the NISP bands, and 27.1-27.3 AB in the VIS band. Using the Lyman-break method in combination with photometric redshifts, we identify $30$ Lyman-break galaxy (LBG) candidates at $z>6$ and 139 extremely red sources (ERSs), most likely at lower redshift. The deeper VIS imaging compared to NISP means we can routinely identify high-redshift Lyman breaks of the order of $3$ magnitudes, which reduces contamination by brown dwarf stars and low-redshift galaxies. Spectroscopic follow-up campaigns of such bright sources will help constrain both the bright end of the ultraviolet galaxy luminosity function and the quasar luminosity function at $z>6$, and constrain the physical nature of these objects. Additionally, we have performed a combined strong lensing and weak lensing analysis of A2390, and demonstrate how Euclid will contribute to better constraining the virial mass of galaxy clusters. From these data, we also identify optical and near-infrared counterparts of known $z>0.6$ clusters, which exhibit strong lensing features, establishing the ability of Euclid to characterize high-redshift clusters. Finally, we provide a glimpse of Euclid's ability to map the intracluster light out to larger radii than current facilities, enabling a better understanding of the cluster assembly history and mapping of the dark matter distribution. This initial dataset illustrates the diverse spectrum of legacy science that will be enabled by the Euclid survey., Comment: Paper submitted as part of the A&A special issue `Euclid on Sky', which contains Euclid key reference papers and first results from the Euclid Early Release Observations. 17 pages, 12 figures

We study the intracluster light (ICL) and intracluster globular clusters (ICGCs) in the nearby Perseus galaxy cluster using Euclid's EROs. By modelling the isophotal and iso-density contours, we mapped the distributions and properties of the ICL and ICGCs out to radii of 200-600 kpc (up to ~1/3 of the virial radius) from the brightest cluster galaxy (BCG). We find that the central 500 kpc hosts 70000$\pm$2800 GCs and $1.7\times10^{12}$ L$_\odot$ of diffuse light from the BCG+ICL in the near-infrared H$_E$. This accounts for 38$\pm$6% of the cluster's total stellar luminosity within this radius. The ICL and ICGCs share a coherent spatial distribution, suggesting a common origin or that a common potential governs their distribution. Their contours on the largest scales (>200 kpc) are offset from the BCG's core westwards by 60 kpc towards several luminous cluster galaxies. This offset is opposite to the displacement observed in the gaseous intracluster medium. The radial surface brightness profile of the BCG+ICL is best described by a double S\'ersic model, with 68$\pm$4% of the H$_E$ light in the extended, outer component. The transition between these components occurs at ~60 kpc, beyond which the isophotes become increasingly elliptical and off-centred. The radial ICGC number density profile closely follows the BCG+ICL profile only beyond this 60 kpc radius, where we find an average of 60-80 GCs per $10^9$ M$_\odot$ of diffuse stellar mass. The BCG+ICL colour becomes increasingly blue with radius, consistent with the stellar populations in the ICL having subsolar metallicities [Fe/H] ~ -0.6 to -1.0. The colour of the ICL, and the specific frequency and luminosity function of the ICGCs suggest that the ICL+ICGCs were tidally stripped from the outskirts of massive satellites with masses of a few $\times10^{10}$ M$_\odot$, with an increasing contribution from dwarf galaxies at large radii., Comment: Paper submitted as part of the A&A special issue `Euclid on Sky', which contains Euclid key reference papers and first results from the Euclid Early Release Observations. 27 pages, 21 figures, 5 tables. Accepted for publication in A&A

We make use of the unprecedented depth, spatial resolution, and field of view of the Euclid Early Release Observations of the Perseus galaxy cluster to detect and characterise the dwarf galaxy population in this massive system. The Euclid high resolution VIS and combined VIS+NIR colour images were visually inspected and dwarf galaxy candidates were identified. Their morphologies, the presence of nuclei, and their globular cluster (GC) richness were visually assessed, complementing an automatic detection of the GC candidates. Structural and photometric parameters, including Euclid filter colours, were extracted from 2-dimensional fitting. Based on this analysis, a total of 1100 dwarf candidates were found across the image, with 638 appearing to be new identifications. The majority (96%) are classified as dwarf ellipticals, 53% are nucleated, 26% are GC-rich, and 6% show disturbed morphologies. A relatively high fraction of galaxies, 8%, are categorised as ultra-diffuse galaxies. The majority of the dwarfs follow the expected scaling relations. Globally, the GC specific frequency, S_N, of the Perseus dwarfs is intermediate between those measured in the Virgo and Coma clusters. While the dwarfs with the largest GC counts are found throughout the Euclid field of view, those located around the east-west strip, where most of the brightest cluster members are found, exhibit larger S_N values, on average. The spatial distribution of the dwarfs, GCs, and intracluster light show a main iso-density/isophotal centre displaced to the west of the bright galaxy light distribution. The ERO imaging of the Perseus cluster demonstrates the unique capability of Euclid to concurrently detect and characterise large samples of dwarfs, their nuclei, and their GC systems, allowing us to construct a detailed picture of the formation and evolution of galaxies over a wide range of mass scales and environments., Comment: 44 pages, 24 figures, 5 tables, paper submitted to A&A as part of the A&A special issue `Euclid on Sky', which contains Euclid key reference papers and first results from the Euclid Early Release Observations

The Euclid ERO programme targeted the Perseus cluster of galaxies, gathering deep data in the central region of the cluster over 0.7 square degree, corresponding to approximately 0.25 r_200. The data set reaches a point-source depth of IE=28.0 (YE, JE, HE = 25.3) AB magnitudes at 5 sigma with a 0.16" and 0.48" FWHM, and a surface brightness limit of 30.1 (29.2) mag per square arcsec. The exceptional depth and spatial resolution of this wide-field multi-band data enable the simultaneous detection and characterisation of both bright and low surface brightness galaxies, along with their globular cluster systems, from the optical to the NIR. This study advances beyond previous analyses of the cluster and enables a range of scientific investigations summarised here. We derive the luminosity and stellar mass functions (LF and SMF) of the Perseus cluster in the Euclid IE band, thanks to supplementary u,g,r,i,z and Halpha data from the CFHT. We adopt a catalogue of 1100 dwarf galaxies, detailed in the corresponding ERO paper. We identify all other sources in the Euclid images and obtain accurate photometric measurements using AutoProf or AstroPhot for 138 bright cluster galaxies, and SourceExtractor for half a million compact sources. Cluster membership for the bright sample is determined by calculating photometric redshifts with Phosphoros. Our LF and SMF are the deepest recorded for the Perseus cluster, highlighting the groundbreaking capabilities of the Euclid telescope. Both the LF and SMF fit a Schechter plus Gaussian model. The LF features a dip at M(IE)=-19 and a faint-end slope of alpha_S = -1.2 to -1.3. The SMF displays a low-mass-end slope of alpha_S = -1.2 to -1.35. These observed slopes are flatter than those predicted for dark matter halos in cosmological simulations, offering significant insights for models of galaxy formation and evolution., Comment: Submitted to A&A, 44 pages, 35 figures, Part of the A&A special issue `Euclid on Sky', which contains Euclid key reference papers and first results from the Euclid Early Release Observations

The Euclid ERO showcase Euclid's capabilities in advance of its main mission, targeting 17 astronomical objects, from galaxy clusters, nearby galaxies, globular clusters, to star-forming regions. A total of 24 hours observing time was allocated in the early months of operation, engaging the scientific community through an early public data release. We describe the development of the ERO pipeline to create visually compelling images while simultaneously meeting the scientific demands within months of launch, leveraging a pragmatic, data-driven development strategy. The pipeline's key requirements are to preserve the image quality and to provide flux calibration and photometry for compact and extended sources. The pipeline's five pillars are: removal of instrumental signatures; astrometric calibration; photometric calibration; image stacking; and the production of science-ready catalogues for both the VIS and NISP instruments. We report a PSF with a full width at half maximum of 0.16" in the optical and 0.49" in the three NIR bands. Our VIS mean absolute flux calibration is accurate to about 1%, and 10% for NISP due to a limited calibration set; both instruments have considerable colour terms. The median depth is 25.3 and 23.2 AB mag with a SNR of 10 for galaxies, and 27.1 and 24.5 AB mag at an SNR of 5 for point sources for VIS and NISP, respectively. Euclid's ability to observe diffuse emission is exceptional due to its extended PSF nearly matching a pure diffraction halo, the best ever achieved by a wide-field, high-resolution imaging telescope. Euclid offers unparalleled capabilities for exploring the LSB Universe across all scales, also opening a new observational window in the NIR. Median surface-brightness levels of 29.9 and 28.3 AB mag per square arcsec are achieved for VIS and NISP, respectively, for detecting a 10 arcsec x 10 arcsec extended feature at the 1 sigma level., Comment: Submitted to A&A, 44 pages, 36 figures - Part of the A&A special issue `Euclid on Sky', which contains Euclid key reference papers and first results from the Euclid Early Release Observations

There exist endless examples of dynamical systems with vast available data and unsatisfying mathematical descriptions. Sparse regression applied to symbolic libraries has quickly emerged as a powerful tool for learning governing equations directly from data; these learned equations balance quantitative accuracy with qualitative simplicity and human interpretability. Here, I present a general purpose, model agnostic sparse regression algorithm that extends a recently proposed exhaustive search leveraging iterative Singular Value Decompositions (SVD). This accelerated scheme, Scalable Pruning for Rapid Identification of Null vecTors (SPRINT), uses bisection with analytic bounds to quickly identify optimal rank-1 modifications to null vectors. It is intended to maintain sensitivity to small coefficients and be of reasonable computational cost for large symbolic libraries. A calculation that would take the age of the universe with an exhaustive search but can be achieved in a day with SPRINT., Comment: Scripts to reproduce all figures are located at https://github.com/mgolden30/SPIDER/tree/main/modified_KS Standalone sparse regression scripts can be found at https://github.com/mgolden30/FastSparseRegression

Many clinical tasks require an understanding of specialized data, such as medical images and genomics, which is not typically found in general-purpose large multimodal models. Building upon Gemini's multimodal models, we develop several models within the new Med-Gemini family that inherit core capabilities of Gemini and are optimized for medical use via fine-tuning with 2D and 3D radiology, histopathology, ophthalmology, dermatology and genomic data. Med-Gemini-2D sets a new standard for AI-based chest X-ray (CXR) report generation based on expert evaluation, exceeding previous best results across two separate datasets by an absolute margin of 1% and 12%, where 57% and 96% of AI reports on normal cases, and 43% and 65% on abnormal cases, are evaluated as "equivalent or better" than the original radiologists' reports. We demonstrate the first ever large multimodal model-based report generation for 3D computed tomography (CT) volumes using Med-Gemini-3D, with 53% of AI reports considered clinically acceptable, although additional research is needed to meet expert radiologist reporting quality. Beyond report generation, Med-Gemini-2D surpasses the previous best performance in CXR visual question answering (VQA) and performs well in CXR classification and radiology VQA, exceeding SoTA or baselines on 17 of 20 tasks. In histopathology, ophthalmology, and dermatology image classification, Med-Gemini-2D surpasses baselines across 18 out of 20 tasks and approaches task-specific model performance. Beyond imaging, Med-Gemini-Polygenic outperforms the standard linear polygenic risk score-based approach for disease risk prediction and generalizes to genetically correlated diseases for which it has never been trained. Although further development and evaluation are necessary in the safety-critical medical domain, our results highlight the potential of Med-Gemini across a wide range of medical tasks.

Training large-scale machine learning models poses distinct system challenges, given both the size and complexity of today's workloads. Recently, many organizations training state-of-the-art Generative AI models have reported cases of instability during training, often taking the form of loss spikes. Numeric deviation has emerged as a potential cause of this training instability, although quantifying this is especially challenging given the costly nature of training runs. In this work, we develop a principled approach to understanding the effects of numeric deviation, and construct proxies to put observations into context when downstream effects are difficult to quantify. As a case study, we apply this framework to analyze the widely-adopted Flash Attention optimization. We find that Flash Attention sees roughly an order of magnitude more numeric deviation as compared to Baseline Attention at BF16 when measured during an isolated forward pass. We then use a data-driven analysis based on the Wasserstein Distance to provide upper bounds on how this numeric deviation impacts model weights during training, finding that the numerical deviation present in Flash Attention is 2-5 times less significant than low-precision training.

Over three decades ago the advection-diffusion equation for a steady fluid velocity field was homogenized, leading to a Stieltjes integral representation for the effective diffusivity, which is given in terms of a spectral measure of a compact self-adjoint operator and the P\'eclet number of the fluid flow. This result was recently extended to space-time periodic flows, instead involving an unbounded self-adjoint operator. Pad\'e approximants provide rigorous upper and lower bounds for Stieltjes functions in terms of the moments of the spectral measure. However, with the lack of a method for calculating the moments of the spectral measure for general fluid velocity fields, the utility of this powerful mathematical framework for calculating bounds for the effective diffusivity has not been fully realized. Here we significantly expand the applicability of this framework by providing an iterative method that enables an arbitrary number of moments, hence bounds, to be calculated analytically in closed form for both spatially and space-time periodic flows. The method is demonstrated for periodic flows in two spatial dimensions. The known asymptotic behavior of the effective diffusivity for a steady flow is accurately captured by high order upper and lower bounds, demonstrating the ability of the method to provide accurate estimates for the effective diffusivity for a broad range of parameter values., Comment: 20 pages, 6 figures, and 2 tables

For countably infinite IFSs on $\mathbb R^2$ consisting of affine contractions with diagonal linear parts, we give conditions under which the affinity dimension is an upper bound for the Hausdorff dimension and a lower bound for the lower box-counting dimension. Moreover, we identify a family of countably infinite IFSs for which the Hausdorff and affinity dimension are equal, and which have full dimension spectrum. The corresponding self-affine sets are related to restricted digit sets for signed L\"uroth expansions., Comment: 16 pages, 1 figure with 3 subfigures

Low density plasmas in curved spacetimes, such as those found in accretion flows around black holes, are challenging to model from first principles, owing to the large scale separation between the characteristic scales of the microscopic processes and large mean-free-paths comparable to the system sizes. Kinetic approaches become necessary to capture the relevant physics but lack the dynamic range to model the global characteristics of the systems. In this paper, we develop new covariant guiding center equations of motion for charges in general relativistic spacetimes that are computationally tractable. We decompose the particle motion into a fast gyration, which we integrate analytically and a slow drift of the guiding center, which can be solved numerically. We derive covariant conservation laws for the motions of the guiding centers and show, through a number of limiting cases, that the equations contain all known drift mechanisms. Finally, we present the general relativistic expressions for the various drift velocities in Schwarzschild spacetimes., Comment: Submitted to ApJ

We show that the momentum-dependent scaling exponents of the holographic fermion self-energy of the conformal-to-AdS$_2$ Gubser-Rocha model can describe new findings from angle-resolved photoemission spectroscopy experiments on a single layer (Pb,Bi)$_{2}$Sr$_{2-x}$La$_x$CuO$_{6+\delta}$ copper-oxide. In particular, it was recently observed, in high-precision measurements on constant energy cuts along the nodal direction, that the spectral function departs from the Lorentzian line shape that is expected from the power-law-liquid model of a nodal self-energy, with an imaginary part featureless in momentum as $\Sigma''_{\text{PLL}}(\omega) \propto (\omega^2)^\alpha$. By direct comparison with experimental results, we provide evidence that this departure from either a Fermi liquid or the power-law liquid, resulting in an asymmetry of the spectral function as a function of momentum around the central peak, is captured at low temperature and all dopings by a semi-holographic model that predicts a momentum-dependent scaling exponent in the electron self-energy as $\Sigma(\omega,k) \propto \omega (-\omega^2)^{\alpha (1 - (k - k_F)/k_F) - 1/2}$, with $\hbar k_F$ the Fermi momentum.

The Quantum Alternating Operator Ansatz (QAOA) is a prominent variational quantum algorithm for solving combinatorial optimization problems. Its effectiveness depends on identifying input parameters that yield high-quality solutions. However, understanding the complexity of training QAOA remains an under-explored area. Previous results have given analytical performance guarantees for a small, fixed number of parameters. At the opposite end of the spectrum, barren plateaus are likely to emerge at $\Omega(n)$ parameters for $n$ qubits. In this work, we study the difficulty of training in the intermediate regime, which is the focus of most current numerical studies and near-term hardware implementations. Through extensive numerical analysis of the quality and quantity of local minima, we argue that QAOA landscapes can exhibit a superpolynomial growth in the number of low-quality local minima even when the number of parameters scales logarithmically with $n$. This means that the common technique of gradient descent from randomly initialized parameters is doomed to fail beyond small $n$, and emphasizes the need for good initial guesses of the optimal parameters., Comment: minor updates