Search

The advent of foundation models (FMs) is transforming medical domain. In ophthalmology, RETFound, a retina-specific FM pre-trained sequentially on 1.4 million natural images and 1.6 million retinal images, has demonstrated high adaptability across clinical applications. Conversely, DINOv2, a general-purpose vision FM pre-trained on 142 million natural images, has shown promise in non-medical domains. However, its applicability to clinical tasks remains underexplored. To address this, we conducted head-to-head evaluations by fine-tuning RETFound and three DINOv2 models (large, base, small) for ocular disease detection and systemic disease prediction tasks, across eight standardized open-source ocular datasets, as well as the Moorfields AlzEye and the UK Biobank datasets. DINOv2-large model outperformed RETFound in detecting diabetic retinopathy (AUROC=0.850-0.952 vs 0.823-0.944, across three datasets, all P<=0.007) and multi-class eye diseases (AUROC=0.892 vs. 0.846, P<0.001). In glaucoma, DINOv2-base model outperformed RETFound (AUROC=0.958 vs 0.940, P<0.001). Conversely, RETFound achieved superior performance over all DINOv2 models in predicting heart failure, myocardial infarction, and ischaemic stroke (AUROC=0.732-0.796 vs 0.663-0.771, all P<0.001). These trends persisted even with 10% of the fine-tuning data. These findings showcase the distinct scenarios where general-purpose and domain-specific FMs excel, highlighting the importance of aligning FM selection with task-specific requirements to optimise clinical performance.

The Herschel open-time key program Disc Emission via a Bias-free Reconnaissance in the Infrared and Sub-millimeter (DEBRIS) is an unbiased survey of the nearest ~100 stars for each stellar type A-M observed with a uniform photometric sensitivity to search for cold debris disks around them. The analysis of the Photoconductor Array Camera and Spectrometer (PACS) photometric observations of the 94 DEBRIS M dwarfs of this program is presented in this paper, following upon two companion papers on the DEBRIS A-star and FGK-star subsamples. In the M-dwarf subsample, two debris disks have been detected, around the M3V dwarf GJ581 and the M4V dwarf FomalhautC (LP876-10). This result gives a disk detection rate of 2.1^{+2.7}_{-0.7}% at the 68% confidence level, significantly less than measured for earlier stellar types in the DEBRIS program. However, we show that the survey of the DEBRIS M-dwarf subsample is about ten times shallower than the surveys of the DEBRIS FGK subsamples when studied in the physical parameter space of the disk's fractional dust luminosity versus blackbody radius. Furthermore, had the DEBRIS K-star subsample been observed at the same shallower depth in this parameter space, its measured disk detection rate would have been statistically consistent with the one found for the M-dwarf subsample. Hence, the incidence of debris disks does not appear to drop from the K subsample to the M subsample of the DEBRIS program, when considering disks in the same region of physical parameter space. An alternative explanation is that the only two bright disks discovered in the M-dwarf subsample would not, in fact, be statistically representative of the whole population., Comment: 12 pages, long Table 3 included, accepted for publication in Astronomy & Astrophysics

Polar metals are an intriguing class of materials in which electric polarisation and metallicity can coexist within a single phase. The unique properties of polar metals challenge expectations, making way for the exploration of exotic phenomena such as unconventional magnetism, hyperferroelectric multiferroicity and developing multifunctional devices that can leverage both the materials electric polarization and its asymmetry in the spin conductivity, that arises due to the Rashba effect. Here, via a high pressure single crystal diffraction study, we report the pressure-induced enhancement of polar distortions in such a metal, Ca$_3$Ru$_2$O$_7$. Our DFT calculations highlight that naive assumptions about the linear dependency between polar distortion amplitudes and the magnitude of the Rashba spin splitting may not be generally valid.

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,700 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 27 pages, 6 figures

Many active galaxies harbor powerful relativistic jets, however, the detailed mechanisms of their formation and acceleration remain poorly understood. To investigate the area of jet acceleration and collimation with the highest available angular resolution, we study the innermost region of the bipolar jet in the nearby low-ionization nuclear emission-line region (LINER) galaxy NGC 1052. We combined observations of NGC 1052 taken with VLBA, GMVA, and EHT over one week in the spring of 2017. For the first time, NGC 1052 was detected with the EHT, providing a size of the central region in-between both jet bases of 250 RS (Schwarzschild radii) perpendicular to the jet axes. This size estimate supports previous studies of the jets expansion profile which suggest two breaks of the profile at around 300 RS and 10000 RS distances to the core. Furthermore, we estimated the magnetic field to be 1.25 Gauss at a distance of 22 {\mu}as from the central engine by fitting a synchrotron-self absorption spectrum to the innermost emission feature, which shows a spectral turn-over at about 130 GHz. Assuming a purely poloidal magnetic field, this implies an upper limit on the magnetic field strength at the event horizon of 26000 Gauss, which is consistent with previous measurements. The complex, low-brightness, double-sided jet structure in NGC 1052 makes it a challenge to detect the source at millimeter (mm) wavelengths. However, our first EHT observations have demonstrated that detection is possible up to at least 230 GHz. This study offers a glimpse through the dense surrounding torus and into the innermost central region, where the jets are formed. This has enabled us to finally resolve this region and provide improved constraints on its expansion and magnetic field strength., Comment: 22 pages, 10 figures, published in A&A

The time-variable emission from the accretion flow of Sgr A*, the supermassive black hole at the Galactic Center, has long been examined in the radio-to-mm, near-infrared (NIR), and X-ray regimes of the electromagnetic spectrum. However, until now, sensitivity and angular resolution have been insufficient in the crucial mid-infrared (MIR) regime. The MIRI instrument on JWST has changed that, and we report the first MIR detection of Sgr A*. The detection was during a flare that lasted about 40 minutes, a duration similar to NIR and X-ray flares, and the source's spectral index steepened as the flare ended. The steepening suggests synchrotron cooling is an important process for Sgr A*'s variability and implies magnetic field strengths $\sim$40--70 Gauss in the emission zone. Observations at $1.3~\mathrm{mm}$ with the Submillimeter Array revealed a counterpart flare lagging the MIR flare by $\approx$10 minutes. The observations can be self-consistently explained as synchrotron radiation from a single population of gradually cooling high-energy electrons accelerated through (a combination of) magnetic reconnection and/or magnetized turbulence., Comment: Accepted for publication ApJL

The 2017 observing campaign of the Event Horizon Telescope (EHT) delivered the first very long baseline interferometry (VLBI) images at the observing frequency of 230 GHz, leading to a number of unique studies on black holes and relativistic jets from active galactic nuclei (AGN). In total, eighteen sources were observed: the main science targets, Sgr A* and M87 along with various calibrators. We investigated the morphology of the sixteen AGN in the EHT 2017 data set, focusing on the properties of the VLBI cores: size, flux density, and brightness temperature. We studied their dependence on the observing frequency in order to compare it with the Blandford-K\"onigl (BK) jet model. We modeled the source structure of seven AGN in the EHT 2017 data set using linearly polarized circular Gaussian components and collected results for the other nine AGN from dedicated EHT publications, complemented by lower frequency data in the 2-86 GHz range. Then, we studied the dependences of the VLBI core flux density, size, and brightness temperature on the frequency measured in the AGN host frame. We compared the observations with the BK jet model and estimated the magnetic field strength dependence on the distance from the central black hole. Our results indicate a deviation from the standard BK model, particularly in the decrease of the brightness temperature with the observing frequency. Either bulk acceleration of the jet material, energy transfer from the magnetic field to the particles, or both are required to explain the observations.

We present optical photometric and spectroscopic observations of the peculiar Type Ia supernova ASASSN-20jq/SN 2020qxp. It is a low-luminosity object with a peak absolute magnitude of $M_B=-17.1\pm0.5$ mag. Despite its low luminosity, its post-peak light-curve decline rate ($\Delta m_{15}(B)=1.35\pm0.09$ mag) and color-stretch parameter (sBV>0.82) are similar to normal SNe Ia, making it an outlier in the luminosity-width and luminosity-color-stretch relations. Early light curves suggest a "bump" during the first 1.4 days of explosion. ASASSN-20jq synthesized a low radioactive $^{56}$Ni mass of $0.09\pm0.01M_\odot$. Near-maximum light spectra reveal strong Si II absorption lines, indicating a cooler photosphere than normal SNe Ia, but lack Ti II absorption lines. Unusually strong O I $\lambda$7773 and Ca II near-infrared triplet absorption features are present. Nebular spectra show a strong, narrow forbidden [Ca II] $\lambda\lambda$7291,7324 doublet emission, rarely seen in SNe Ia except in some Type Iax events. Marginal detection of [O I] $\lambda\lambda$6300,6364 doublet emission, which is extremely rare, is observed. Both [Ca II] and [O I] lines are redshifted by $\sim2000$ km/s. A strong [Fe II] $\lambda$7155 emission line with a tilted-top profile, identical to the [Fe II] $\lambda$16433 profile, is also observed. These asymmetric [Fe II] profiles and redshifted [Ca II] and [O I] emissions suggest a high central density white dwarf progenitor undergoing an off-center delayed-detonation explosion mechanism, producing roughly equal amounts of $^{56}$Ni in deflagration and detonation phases. This distinguishes ASASSN-20jq from normal and subluminous SNe Ia. ASASSN-20jq's light curve and spectra do not align with any single SNe Ia subclass but show similarities to 2002es-like objects. Thus, we add it as an extreme candidate within the heterogeneous parameter space of 2002es-like SNe Ia., Comment: 27 pages, 20 figures, 5 tables, submitted to A&A

We report the detection of the [O III] auroral line in 42 galaxies within the redshift range of $3 < z < 10$. These galaxies were selected from publicly available JWST data releases, including the JADES and PRIMALsurveys, and observed using both the low-resolution PRISM/CLEAR configuration and medium-resolution gratings. The measured electron temperatures in the high-ionization regions of these galaxies range from $T_e$([O III]) = 12,000 to 24,000 K, consistent with temperatures observed in local metal-poor galaxies and previous JWST studies. In 10 galaxies, we also detect the [O II] auroral line, allowing us to determine electron temperatures in the low-ionization regions, which range between $T_e$([O II]) = 10,830 and 20,000 K. The direct-$T_e$-based metallicities of our sample span from 12 + log(O/H) = 7.2 to 8.4, indicating these high-redshift galaxies are relatively metal-poor. By combining our sample with 25 galaxies from the literature, we expand the dataset to a total of 67 galaxies within $3 < z < 10$, effectively more than doubling the previous sample size for direct-$T_e$ based metallicity studies. This larger dataset allow us to derive empirical metallicity calibration relations based exclusively on high-redshift galaxies, using six key line ratios: R3, R2, R23, Ne3O2, O32, and O3N2. Notably, we derive a novel metallicity calibration relation for the first time using high-redshift $T_e$-based metallicities: $\hat{R}$ = 0.18log $R2$ + 0.98log $R3$. This new calibration significantly reduces the scatter in high-redshift galaxies compared to the $\hat{R}$ relation previously calibrated for low-redshift galaxies., Comment: 23 pages, 10 figures, submitted to ApJ

Sound decision-making relies on accurate prediction for tangible outcomes ranging from military conflict to disease outbreaks. To improve crowdsourced forecasting accuracy, we developed SAGE, a hybrid forecasting system that combines human and machine generated forecasts. The system provides a platform where users can interact with machine models and thus anchor their judgments on an objective benchmark. The system also aggregates human and machine forecasts weighting both for propinquity and based on assessed skill while adjusting for overconfidence. We present results from the Hybrid Forecasting Competition (HFC) - larger than comparable forecasting tournaments - including 1085 users forecasting 398 real-world forecasting problems over eight months. Our main result is that the hybrid system generated more accurate forecasts compared to a human-only baseline which had no machine generated predictions. We found that skilled forecasters who had access to machine-generated forecasts outperformed those who only viewed historical data. We also demonstrated the inclusion of machine-generated forecasts in our aggregation algorithms improved performance, both in terms of accuracy and scalability. This suggests that hybrid forecasting systems, which potentially require fewer human resources, can be a viable approach for maintaining a competitive level of accuracy over a larger number of forecasting questions., Comment: 20 pages, 6 figures, 4 tables

The Wallace Foundation commissioned Research for Action and McClanahan Associates to study how out-of-school time intermediaries (OSTIs) responded to the COVID-19 pandemic and the role they could play in supporting communities moving forward. Citywide out-of-school time (OST) systems encompass a wide variety of afterschool and summer programs and associated provider organizations, youth participants, and their families. These systems were stretched during the COVID-19 pandemic and the national racial reckoning to respond to community needs in new and more equitable ways. Also impacted were OSTIs that manage OST systems within individual cities. OSTIs operate through various governance structures and perform a range of coordination functions intended to ensure quality, access, and equity in OST systems so that all youth have access to the unique out-of-school resources and opportunities they need to thrive.

Drawing on qualitative data collected in a sample of colleges as part of a larger study on the implementation and impact of Assembly Bill 705 in California, this paper explores the rollout of corequisite reforms, focusing on the use of embedded tutors in introductory math and English courses as a strategy to meet to the needs of students. This paper highlights promising practices identified through extant research and fieldwork at study institutions, provides additional evidence on the value of the reform, discusses challenges, and makes recommendations for the field.

Research for Action (RFA) in partnership with the University of Texas at Austin is engaged in a five-year mixed-methods study of the reforms associated with California AB 705. Over the course of the study, our team will assess the implementation, impact, and cost effectiveness of reforms associated with the law. This second interim report, presented at the conclusion of year three of the study, focuses on gaining a deeper understanding of on-campus implementation through a faculty survey administered to math and English departments across our study sample, an Interrupted Time Series analysis with nine cohorts of FTIC student data, and preliminary data collection for our cost effectiveness study. Collectively, these data highlight significant changes that colleges have made on campus regarding shifting enrollments from developmental education into transfer-level coursework in both English and math, and providing additional supports to students to promote retention and completion. We find that AB 705 has demonstrated notable successes in improving enrollment and completion rates in transfer-level courses, particularly in math, among FTIC students in California's community colleges. While our survey results suggest that faculty believe additional resources and supports would be helpful, most faculty report that implementation supports have been adequate.

A major challenge for reinforcement learning is automatically generating curricula to reduce training time or improve performance in some target task. We introduce SEBNs (Skill-Environment Bayesian Networks) which model a probabilistic relationship between a set of skills, a set of goals that relate to the reward structure, and a set of environment features to predict policy performance on (possibly unseen) tasks. We develop an algorithm that uses the inferred estimates of agent success from SEBN to weigh the possible next tasks by expected improvement. We evaluate the benefit of the resulting curriculum on three environments: a discrete gridworld, continuous control, and simulated robotics. The results show that curricula constructed using SEBN frequently outperform other baselines.

As large language models (LLMs) become integrated into everyday applications, ensuring their robustness and security is increasingly critical. In particular, LLMs can be manipulated into unsafe behaviour by prompts known as jailbreaks. The variety of jailbreak styles is growing, necessitating the use of external defences known as guardrails. While many jailbreak defences have been proposed, not all defences are able to handle new out-of-distribution attacks due to the narrow segment of jailbreaks used to align them. Moreover, the lack of systematisation around defences has created significant gaps in their practical application. In this work, we perform systematic benchmarking across 15 different defences, considering a broad swathe of malicious and benign datasets. We find that there is significant performance variation depending on the style of jailbreak a defence is subject to. Additionally, we show that based on current datasets available for evaluation, simple baselines can display competitive out-of-distribution performance compared to many state-of-the-art defences. Code is available at https://github.com/IBM/Adversarial-Prompt-Evaluation., Comment: NeurIPS 2024, Safe Generative AI Workshop

Low Earth orbit (LEO) satellite constellations bring broadband internet and cellular service to the most remote locations on the planet. Unfortunately, many of these locations also host some of the world's best optical and radio astronomy (RA) observatories. With the number of LEO satellites expected to increase by an order of magnitude in the upcoming decade, satellite downlink radio frequency interference (RFI) is a growing concern in protected radio-quiet areas like the United States National Radio Quiet Zone. When these satellites transmit in the spectrum near protected RA bands, undesired out-of-band emission can leak into these protected bands and impact scientific observations. In this paper, we present a self-reporting system - Operational Data Sharing (ODS) - which enables mutual awareness by publishing radio telescopes' operational information to a protected database that is available to satellite operators through a representational state transfer application programming interface (REST API). Satellite operators can use the ODS data to adapt their downlink tasking algorithms in real time to avoid overwhelming sensitive RA facilities, particularly, through the novel Telescope Boresight Avoidance (TBA) technique. Preliminary results from recent experiments between the NRAO and the SpaceX Starlink teams demonstrate the effectiveness of the ODS and TBA in reducing downlink RFI in the Karl G. Jansky Very Large Array's observations in the 1990-1995 MHz and 10.7-12.7 GHz bands. This automated ODS system is beginning to be implemented by other RA facilities and could be utilized by other satellite operators in the near future., Comment: 9 pages, 6 figures, draft version, submitted for review to the Astrophysical Journal Letters (ApJL) on 20 Feb 2025

The goal of the AlgoPerf: Training Algorithms competition is to evaluate practical speed-ups in neural network training achieved solely by improving the underlying training algorithms. In the external tuning ruleset, submissions must provide workload-agnostic hyperparameter search spaces, while in the self-tuning ruleset they must be completely hyperparameter-free. In both rulesets, submissions are compared on time-to-result across multiple deep learning workloads, training on fixed hardware. This paper presents the inaugural AlgoPerf competition's results, which drew 18 diverse submissions from 10 teams. Our investigation reveals several key findings: (1) The winning submission in the external tuning ruleset, using Distributed Shampoo, demonstrates the effectiveness of non-diagonal preconditioning over popular methods like Adam, even when compared on wall-clock runtime. (2) The winning submission in the self-tuning ruleset, based on the Schedule Free AdamW algorithm, demonstrates a new level of effectiveness for completely hyperparameter-free training algorithms. (3) The top-scoring submissions were surprisingly robust to workload changes. We also discuss the engineering challenges encountered in ensuring a fair comparison between different training algorithms. These results highlight both the significant progress so far, and the considerable room for further improvements., Comment: ICLR 2025; 23 pages, 5 figures, 8 tables

Recent copyright agreements between AI companies and content creators have highlighted the need for precise control over language models' ability to reproduce copyrighted content. While existing approaches rely on either complete concept removal through unlearning or simple output filtering, we propose Obliviate, a novel post-training technique that selectively prevents verbatim reproduction of specific text while preserving semantic understanding. Obliviate operates by selecting tokens within memorized sequences and modifying the model's probability distribution to prevent exact reproduction while maintaining contextual understanding. We evaluate Obliviate on multiple large language models (LLaMA-3.1 8B, LLaMA-3.1-instruct 8B, Qwen-2.5-7B, and Yi-1.5 6B) across both synthetic memorization tasks and organic copyright content. Our results demonstrate that Obliviate achieves orders of magnitude reduction, e.g., 100x, in verbatim memorization while maintaining model performance within 1% of baseline on standard benchmarks (HellaSwag, MMLU, TruthfulQA, and Winogrande). This makes Obliviate particularly suitable for practical deployment scenarios where companies need to efficiently address copyright concerns in pretrained models without compromising their general capabilities.

We present an analysis of the magnetic field strength and morphology in the Sagittarius C complex (Sgr C; G359.43-0.09) in the Milky Way Galaxy's Central Molecular Zone (CMZ) using the 214 $\mu$m polarimetry data acquired with the High-Angular-Resolution Wideband Camera+ (HAWC+) instrument aboard the Stratospheric Observatory for Infrared Astronomy (SOFIA). We use several hundred magnetic field pseudovectors in the Sgr C region to trace the projected magnetic field orientation within cold molecular gas clouds, and as is the trend throughout the CMZ, they show a higher polarization fraction toward the periphery of the clouds. We conduct a modified Davis-Chandrasekhar-Fermi (DCF) analysis of individual clouds and find that the sky-plane magnetic field strength varies from highly turbulent regions having inferred strengths of $\sim30~\mu{\rm G}$ to regions of relatively uniform field orientation having strengths of $\sim 1~{\rm mG}$. The magnetic field orientations suggest that outflows from active star-forming regions, such as the extended green object (EGO) G359.43-0.10 and the protostellar source FIR-4 (G359.43+0.02), cause high turbulence in their vicinity. The magnetic field direction is found to be tangential to the surface of the Sgr C HII region, as well as two [CII] emission cavities around this region. Several other features in the vicinity of Sgr C, especially numerous non-thermal filaments (NTFs) and a diffuse source of X-ray emission towards the southwest of the \hii{} region, are discussed with regard to the observed magnetic field pseudovectors., Comment: 30 pages, 18 figures. Submitted to ApJ

Reasoning about images with rich text, such as charts and documents, is a critical application of vision-language models (VLMs). However, VLMs often struggle in these domains due to the scarcity of diverse text-rich vision-language data. To address this challenge, we present CoSyn, a framework that leverages the coding capabilities of text-only large language models (LLMs) to automatically create synthetic text-rich multimodal data. Given input text describing a target domain (e.g., "nutrition fact labels"), CoSyn prompts an LLM to generate code (Python, HTML, LaTeX, etc.) for rendering synthetic images. With the underlying code as textual representations of the synthetic images, CoSyn can generate high-quality instruction-tuning data, again relying on a text-only LLM. Using CoSyn, we constructed a dataset comprising 400K images and 2.7M rows of vision-language instruction-tuning data. Comprehensive experiments on seven benchmarks demonstrate that models trained on our synthetic data achieve state-of-the-art performance among competitive open-source models, including Llama 3.2, and surpass proprietary models such as GPT-4V and Gemini 1.5 Flash. Furthermore, CoSyn can produce synthetic pointing data, enabling VLMs to ground information within input images, showcasing its potential for developing multimodal agents capable of acting in real-world environments., Comment: 20 pages, 19 figures, 9 tables, website: https://yueyang1996.github.io/cosyn/

Amorphous Al$_{2}$O$_{3}$ is a fundamental component of modern superconducting qubits. While amphorphous oxides offer distinct advantages, such as directional isotropy and a consistent bulk electronic gap, in realistic systems these compounds support two-level systems (TLSs) which couple to the qubit, expediting decoherence. In this work, we perform a first-principles study of amorphous Al$_{2}$O$_{3}$ and identify low-energy modes in the electronic and phonon spectra as a possible origin for TLSs., Comment: 5+1 Pages, 7 + 3 Figures

In using the QAOA algorithm for the MaxCut problem one encodes the number of edges connecting the sets resulting from a partition of the vertices of a graph into phases of amplitudes of a quantum state (QAOA state). One wants to use the quantum state to find partitions with large numbers of edges connecting them. In the QAOA algorithm this is done by using a mixing operation and parameter optimization. Here we want to see what can be done if we only use simple aspects of quantum mechanics, interference and measurement, to extract information from the QAOA state. The idea is to use constructive interference to enhance the amplitudes corresponding to partitions with a large number of edges between the sets. We examine examples, but analytically and numerically. We also show how the results of sequences of measurements can be used to gain information about the landscape of solutions.

Coastal regions are particularly vulnerable to the impacts of rising sea levels and extreme weather events. Accurate real-time forecasting of hydrodynamic processes in these areas is essential for infrastructure planning and climate adaptation. In this study, we present the Multiple-Input Temporal Operator Network (MITONet), a novel autoregressive neural emulator that employs dimensionality reduction to efficiently approximate high-dimensional numerical solvers for complex, nonlinear problems that are governed by time-dependent, parameterized partial differential equations. Although MITONet is applicable to a wide range of problems, we showcase its capabilities by forecasting regional tide-driven dynamics described by the two-dimensional shallow-water equations, while incorporating initial conditions, boundary conditions, and a varying domain parameter. We demonstrate MITONet's performance in a real-world application, highlighting its ability to make accurate predictions by extrapolating both in time and parametric space.

We consider machine learning models, learned from data, to be an important, intensional, kind of data in themselves. As such, various analysis tasks on models can be thought of as queries over this intensional data, often combined with extensional data such as data for training or validation. We demonstrate that relational database systems and SQL can actually be well suited for many such tasks.

We have compiled the $\sim$4-400 GHz broad spectra of 32 Class II protoplanetary disks in the Taurus-Auriga region, which represents the brightest one-third of sources detected in the submillimeter band in this region. The spectra at >20 GHz frequency can be described with a piecewise function: (1) a power law with a spectral index $\sim$2 at >200 GHz, (2) a power law with spectral index in the range 0.3-4.2 at 20-50 GHz, and (3) a transition region in between these two power laws which can be characterized by a sigmoid function. This suggests that the flux densities at >200 GHz and <50 GHz are dominated by distinct emission components. At >200 GHz, the emission is likely dominated by the optically thick dust thermal emission in the bulk of the disks. In some sources that were not detected at 6.8 GHz or 10 GHz, embedded high-density dust substructures may contribute to a significant fraction of the flux densities at 30-50 GHz, and the spectral indices are mostly consistent with 2.0. Although, at 30-50 GHz, free-free and/or synchrotron emission may be significant, and some sources in our sample have spectral indices < 2.0. Based on these results, we hypothesize that high-density dust substructures (e.g., vortices) are often found in resolved Class II protoplanetary disks, and are a precursor to the formation of kilometer-sized planetesimals and rocky planets. They may not present high contrast at >200 GHz frequencies owing to the high optical depth. To probe these dust substructures, high angular resolution observations at <100 GHz are necessary to distinguish them from free-free and synchrotron emission sources. Otherwise, in the analyses of the spatially unresolved spectra, one needs to simultaneously constrain the flux densities of free-free, synchrotron, and dust emission with the observations at $\sim$5-50 GHz., Comment: Accepted for publication in ApJS

Although individual neurons and neural populations exhibit the phenomenon of representational drift, perceptual and behavioral outputs of many neural circuits can remain stable across time scales over which representational drift is substantial. These observations motivate a dynamical systems framework for neural network activity that focuses on the concept of \emph{latent processing units,} core elements for robust coding and computation embedded in collective neural dynamics. Our theoretical treatment of these latent processing units yields five key attributes of computing through neural network dynamics. First, neural computations that are low-dimensional can nevertheless generate high-dimensional neural dynamics. Second, the manifolds defined by neural dynamical trajectories exhibit an inherent coding redundancy as a direct consequence of the universal computing capabilities of the underlying dynamical system. Third, linear readouts or decoders of neural population activity can suffice to optimally subserve downstream circuits controlling behavioral outputs. Fourth, whereas recordings from thousands of neurons may suffice for near optimal decoding from instantaneous neural activity patterns, experimental access to millions of neurons may be necessary to predict neural ensemble dynamical trajectories across timescales of seconds. Fifth, despite the variable activity of single cells, neural networks can maintain stable representations of the variables computed by the latent processing units, thereby making computations robust to representational drift. Overall, our framework for latent computation provides an analytic description and empirically testable predictions regarding how large systems of neurons perform robust computations via their collective dynamics.

Vehicle-to-everything (V2X) collaborative perception has emerged as a promising solution to address the limitations of single-vehicle perception systems. However, existing V2X datasets are limited in scope, diversity, and quality. To address these gaps, we present Mixed Signals, a comprehensive V2X dataset featuring 45.1k point clouds and 240.6k bounding boxes collected from three connected autonomous vehicles (CAVs) equipped with two different types of LiDAR sensors, plus a roadside unit with dual LiDARs. Our dataset provides precisely aligned point clouds and bounding box annotations across 10 classes, ensuring reliable data for perception training. We provide detailed statistical analysis on the quality of our dataset and extensively benchmark existing V2X methods on it. Mixed Signals V2X Dataset is one of the highest quality, large-scale datasets publicly available for V2X perception research. Details on the website https://mixedsignalsdataset.cs.cornell.edu/.

Concept drift refers to gradual or sudden changes in the properties of data that affect the accuracy of machine learning models. In this paper, we address the problem of concept drift detection in the malware domain. Specifically, we propose and analyze a clustering-based approach to detecting concept drift. Using a subset of the KronoDroid dataset, malware samples are partitioned into temporal batches and analyzed using MiniBatch $K$-Means clustering. The silhouette coefficient is used as a metric to identify points in time where concept drift has likely occurred. To verify our drift detection results, we train learning models under three realistic scenarios, which we refer to as static training, periodic retraining, and drift-aware retraining. In each scenario, we consider four supervised classifiers, namely, Multilayer Perceptron (MLP), Support Vector Machine (SVM), Random Forest, and XGBoost. Experimental results demonstrate that drift-aware retraining guided by silhouette coefficient thresholding achieves classification accuracy far superior to static models, and generally within 1% of periodic retraining, while also being far more efficient than periodic retraining. These results provide strong evidence that our clustering-based approach is effective at detecting concept drift, while also illustrating a highly practical and efficient fully automated approach to improved malware classification via concept drift detection.

The Simulating eXtreme Spacetimes Collaboration's code SpEC can now routinely simulate binary black hole mergers undergoing $\sim25$ orbits, with the longest simulations undergoing nearly $\sim180$ orbits. While this sounds impressive, the mismatch between the highest resolutions for this long simulation is $\mathcal{O}(10^{-1})$. Meanwhile, the mismatch between resolutions for the more typical simulations tends to be $\mathcal{O}(10^{-4})$, despite the resolutions being similar to the long simulations'. In this note, we explain why mismatch alone gives an incomplete picture of code -- and waveform -- quality, especially in the context of providing waveform templates for LISA and 3G detectors, which require templates with $\mathcal{O}(10^{3}) - \mathcal{O}(10^{5})$ orbits. We argue that to ready the GW community for the sensitivity of future detectors, numerical relativity groups must be aware of this caveat, and also run future simulations with at least three resolutions to properly assess waveform accuracy., Comment: 8 pages, 2 figures

The process tensor provides a general representation of a quantum system evolving under repeated interventions and is fundamental for numerical simulations of local many-body dynamics. In this work, we introduce the projected process ensemble, an ensemble of pure output states of a process tensor in a given basis of local interventions, and use it to define increasingly more fine-grained probes of quantum chaos. The first moment of this ensemble encapsulates numerous previously studied chaos quantifiers, including the Alicki-Fannes quantum dynamical entropy, butterfly flutter fidelity, and spatiotemporal entanglement. We discover characteristic entanglement structures within the ensemble's higher moments that can sharply distinguish chaotic from integrable dynamics, overcoming deficiencies of the quantum dynamical and spatiotemporal entropies. These conclusions are supported by extensive numerical simulations of many-body dynamics for a range of spin-chain models, including non-interacting, interacting-integrable, chaotic, and many-body localized regimes. Our work elucidates the fingerprints of chaos on spatiotemporal correlations in quantum stochastic processes, and provides a unified framework for analyzing the complexity of unitary and monitored many-body dynamics.

Global challenges such as food supply chain disruptions, public health crises, and natural hazard responses require access to and integration of diverse datasets, many of which are geospatial. Over the past few years, a growing number of (geo)portals have been developed to address this need. However, most existing (geo)portals are stacked by separated or sparsely connected data "silos" impeding effective data consolidation. A new way of sharing and reusing geospatial data is therefore urgently needed. In this work, we introduce KnowWhereGraph, a knowledge graph-based data integration, enrichment, and synthesis framework that not only includes schemas and data related to human and environmental systems but also provides a suite of supporting tools for accessing this information. The KnowWhereGraph aims to address the challenge of data integration by building a large-scale, cross-domain, pre-integrated, FAIR-principles-based, and AI-ready data warehouse rooted in knowledge graphs. We highlight the design principles of KnowWhereGraph, emphasizing the roles of space, place, and time in bridging various data "silos". Additionally, we demonstrate multiple use cases where the proposed geospatial knowledge graph and its associated tools empower decision-makers to uncover insights that are often hidden within complex and poorly interoperable datasets.

Alignment techniques have become central to ensuring that Large Language Models (LLMs) generate outputs consistent with human values. However, existing alignment paradigms often model an averaged or monolithic preference, failing to account for the diversity of perspectives across cultures, demographics, and communities. This limitation is particularly critical in health-related scenarios, where plurality is essential due to the influence of culture, religion, personal values, and conflicting opinions. Despite progress in pluralistic alignment, no prior work has focused on health, likely due to the unavailability of publicly available datasets. To address this gap, we introduce VITAL, a new benchmark dataset comprising 13.1K value-laden situations and 5.4K multiple-choice questions focused on health, designed to assess and benchmark pluralistic alignment methodologies. Through extensive evaluation of eight LLMs of varying sizes, we demonstrate that existing pluralistic alignment techniques fall short in effectively accommodating diverse healthcare beliefs, underscoring the need for tailored AI alignment in specific domains. This work highlights the limitations of current approaches and lays the groundwork for developing health-specific alignment solutions., Comment: Under review

The rise of Large Language Models (LLMs) has led to significant interest in prompt compression, a technique aimed at reducing the length of input prompts while preserving critical information. However, the prominent approaches in prompt compression often require explicit questions or handcrafted templates for compression, limiting their generalizability. We propose Task-agnostic Prompt Compression (TPC), a novel framework that generalizes compression across tasks and domains without requiring input questions or templates. TPC generates a context-relevant task description using a task descriptor trained on a curated dataset of context and query pairs, and fine-tuned via reinforcement learning with a reward function designed to capture the most relevant information. The task descriptor is then utilized to compute the relevance of each sentence in the prompt to generate the compressed prompt. We introduce 3 model sizes (Base, Large, and Huge), where the largest model outperforms the existing state-of-the-art methods on LongBench and ZeroSCROLLS benchmarks, and our smallest model performs comparable to the existing solutions while being considerably smaller.

Trap-assisted nonradiative recombination has been shown to limit the efficiency of optoelectronic devices. While substitutional carbon ($\mathrm{C_N}$) has been suggested to be a nonradiative recombination center in GaN devices, a complete recombination cycle including the two charge-state transition levels has not been previously described. In this work, we investigate the trap-assisted recombination process due to $\mathrm{C_N}$ in GaN, including multiphonon emission (MPE), radiative recombination, trap-assisted Auger-Meitner (TAAM) recombination, as well as thermal emission of holes. Our study shows the key role of TAAM processes at the high carrier densities relevant for devices. We also reveal the carrier-density regimes where thermal emission and radiative recombination are expected to play an observable role. Our results highlight that carbon concentrations exceeding $\sim$10$^{17}$ cm$^{-3}$ can have a noticeable impact on device efficiency, not just in GaN active layers but also in InGaN and AlGaN. Our comprehensive formalism not only offers detailed results for carbon but provides a general framework for assessing the multiple processes that participate in trap-assisted recombination in semiconductors.

The paper proposes a novel machine learning-based approach to the pathfinding problem on extremely large graphs. This method leverages diffusion distance estimation via a neural network and uses beam search for pathfinding. We demonstrate its efficiency by finding solutions for 4x4x4 and 5x5x5 Rubik's cubes with unprecedentedly short solution lengths, outperforming all available solvers and introducing the first machine learning solver beyond the 3x3x3 case. In particular, it surpasses every single case of the combined best results in the Kaggle Santa 2023 challenge, which involved over 1,000 teams. For the 3x3x3 Rubik's cube, our approach achieves an optimality rate exceeding 98%, matching the performance of task-specific solvers and significantly outperforming prior solutions such as DeepCubeA (60.3%) and EfficientCube (69.6%). Additionally, our solution is more than 26 times faster in solving 3x3x3 Rubik's cubes while requiring up to 18.5 times less model training time than the most efficient state-of-the-art competitor., Comment: 12 pages, 3 tables, 3 figures

We present the discovery of the variable optical counterpart to PSR J2055+1545, a redback millisecond pulsar, and the first radial velocity curve of its companion star. The multi-band optical light curves of this system show a $0.4$$-$$0.6 \ \mathrm{mag}$ amplitude modulation with a single peak per orbit and variable colours, suggesting that the companion is mildly irradiated by the pulsar wind. We find that the flux maximum is asymmetric and occurs at orbital phase $\simeq0.4$, anticipating the superior conjunction of the companion (where the optical emission of irradiated redback companions is typically brightest). We ascribe this asymmetry, well fit with a hot spot in our light curve modelling, to irradiation from the intrabinary shock between pulsar and companion winds. The optical spectra obtained with the \textit{Gran Telescopio Canarias} reveal a G-dwarf companion star with temperatures of $5749 \pm 34 \ \mathrm{K}$ and $6106 \pm 35 \ \mathrm{K}$ at its inferior and superior orbital conjunctions, respectively, and a radial velocity semi-amplitude of $385 \pm 3 \ \mathrm{km}\ \mathrm{s}^{-1}$. Our best-fit model yields a neutron star mass of $1.7^{+0.4}_{-0.1} \ \mathrm{M_{sun}}$ and a companion mass of $0.29^{+0.07}_{-0.01} \ \mathrm{M_{sun}}$. Based on the close similarity between the optical light curve of PSR~J2055$+$1545 and those observed from PSR J1023+0038 and PSR J1227-4853 during their rotation-powered states, we suggest this system may develop an accretion disc in the future and manifest as a transitional millisecond pulsar., Comment: 15 pages, 9 figures, 3 tables. This paper has been accepted for publication in ApJ, and currently in press

We present initial results from extremely well-resolved 3D magnetohydrodynamical simulations of idealized galaxy clusters, conducted using the AthenaPK code on the Frontier exascale supercomputer. These simulations explore the self-regulation of galaxy groups and cool-core clusters by cold gas-triggered active galactic nucleus (AGN) feedback incorporating magnetized kinetic jets. Our simulation campaign includes simulations of galaxy groups and clusters with a range of masses and intragroup and intracluster medium properties. In this paper we present results that focus on a Perseus-like cluster. We find that the simulated clusters are self-regulating, with the cluster cores staying at a roughly constant thermodynamic state and AGN jet power staying at physically reasonable values ($\simeq 10^{44}-10^{45}$~erg/s) for billions of years without a discernible duty cycle. These simulations also produce significant amounts of cold gas, with calculations having strong magnetic fields generally both promoting cold gas formation and allowing cold gas out to much larger clustercentric radii ($\simeq 100$~kpc) than simulations with weak or no fields ($\simeq 10$~kpc), and also having more filamentary cold gas morphology. We find that AGN feedback significantly increases the strength of magnetic fields at the center of the cluster. We also find that the magnetized turbulence generated by the AGN results in turbulence where the velocity power spectra are tied to AGN activity whereas the magnetic energy spectra are much less impacted after reaching a stationary state., Comment: 25 pages, 14 figures. Submitted to ApJ. Comments welcome. More info/material: https://xmagnet-simulations.github.io

Most heavy computation occurs on servers owned by a second party. This reduces data privacy, resulting in interest in data-oblivious computation, which typically severely degrades performance. Secure and fast remote computation is particularly important for linear algebra, which comprises a large fraction of total computation and is best run on highly specialized hardware often only accessible through the cloud. We state the natural efficiency and security desiderata for fast, remote, and data-oblivious linear algebra, conjecture the existence of matrix and vector families implying satisfactory algorithms, and provide such an algorithm contingent on common cryptographic assumptions. We achieve sublinear overhead for the server, dramatically reduced computation cost for the client, and various other practical advantages over previous algorithms. Keywords: Data Privacy, Data-Oblivious Computation, Delegation, Homomorphic Encryption, Cloud Computing, Algorithm Efficiency, Sublinear Overhead, LPN, Matrix Multiplication.

We present a novel reasoning approach called Flow-of-Options (FoO), designed to address intrinsic biases in Large Language Models (LLMs). FoO enables LLMs to systematically explore a diverse range of possibilities in their reasoning, as demonstrated by an FoO-based agentic system for autonomously solving Machine Learning tasks (AutoML). Our framework outperforms state-of-the-art baselines, achieving improvements of 38.2% - 69.2% on standard data science tasks, and 37.4% - 47.9% on therapeutic chemistry tasks. With an overall operation cost under $1 per task, our framework is well-suited for cost-sensitive applications. Beyond classification and regression, we illustrate the broader applicability of our FoO-based agentic system to tasks such as reinforcement learning and image generation. Our framework presents significant advancements compared to current state-of-the-art agentic systems for AutoML, due to the benefits of FoO in enforcing diversity in LLM solutions through compressed, explainable representations that also support long-term memory when combined with case-based reasoning., Comment: Github code: https://github.com/flagshippioneering/Flow-of-Options

With the 4-meter Multi-Object Spectroscopic Telescope (4MOST) expected to provide an influx of transient spectra when it begins observations in early 2026 we consider the potential for real-time classification of these spectra. We investigate three extant spectroscopic transient classifiers: the Deep Automated Supernova and Host classifier (DASH), Next Generation SuperFit (NGSF) and SuperNova IDentification (SNID), with a focus on comparing the efficiency and purity of the transient samples they produce. We discuss our method for simulating realistic, 4MOST-like, host-galaxy contaminated spectra and determining quality cuts for each classifier used to ensure pure SN Ia samples while maintaining efficient classification in other transient classes. We investigate the classifiers individually and in combinations. We find that a combination of DASH and NGSF can produce a SN Ia sample with a purity of 99.9% while successfully classifying 70% of SNe Ia. However, it struggles to classify non-SN Ia transients. We investigate photometric cuts to transient magnitude and transient flux fraction, finding that both can be used to improve transient classification efficiencies by 7--25% depending on the transient subclass. Finally, we present an example classification plan for live classification and the predicted purities and efficiencies across five transient classes: Ia, Ibc, II, superluminous and non-supernova transients., Comment: Submitted to MNRAS

As vehicle automation technology continues to mature, there is a necessity for robust remote monitoring and intervention features. These are essential for intervening during vehicle malfunctions, challenging road conditions, or in areas that are difficult to navigate. This evolution in the role of the human operator - from a constant driver to an intermittent teleoperator - necessitates the development of suitable interaction interfaces. While some interfaces were suggested, a comparative study is missing. We designed, implemented, and evaluated three interaction concepts (path planning, trajectory guidance, and waypoint guidance) with up to four concurrent requests of automated vehicles in a within-subjects study with N=23 participants. The results showed a clear preference for the path planning concept. It also led to the highest usability but lower satisfaction. With trajectory guidance, the fewest requests were resolved. The study's findings contribute to the ongoing development of HMIs focused on the remote assistance of automated vehicles., Comment: to be presented at CHI 2025

Job stability - encompassing secure contracts, adequate wages, social benefits, and career opportunities - is a critical determinant in reducing monetary poverty, as it provides households with reliable income and enhances economic well-being. This study leverages EU-SILC survey and census data to estimate the causal effect of job stability on monetary poverty across Italian provinces, quantifying its influence and analyzing regional disparities. We introduce a novel causal small area estimation (CSAE) framework that integrates global and local estimation strategies for heterogeneous treatment effect estimation, effectively addressing data sparsity at the provincial level. Furthermore, we develop a general bootstrap scheme to construct reliable confidence intervals, applicable regardless of the method used for estimating nuisance parameters. Extensive simulation studies demonstrate that our proposed estimators outperform classical causal inference methods in terms of stability while maintaining computational scalability for large datasets. Applying this methodology to real-world data, we uncover significant relationships between job stability and poverty across six Italian regions, offering critical insights into regional disparities and their implications for evidence-based policy design., Comment: 49 pages, 9 figures, 9 tables

It is common in nature to see aggregation of objects in space. Exploring the mechanism associated with the locations of such clustered observations can be essential to understanding the phenomenon, such as the source of spatial heterogeneity, or comparison to other event generating processes in the same domain. Log-Gaussian Cox processes (LGCPs) represent an important class of models for quantifying aggregation in a spatial point pattern. However, implementing likelihood-based Bayesian inference for such models presents many computational challenges, particularly in high dimensions. In this paper, we propose a novel likelihood-free inference approach for LGCPs using the recently developed BayesFlow approach, where invertible neural networks are employed to approximate the posterior distribution of the parameters of interest. BayesFlow is a neural simulation-based method based on "amortized" posterior estimation. That is, after an initial training procedure, fast feed-forward operations allow rapid posterior inference for any data within the same model family. Comprehensive numerical studies validate the reliability of the framework and show that BayesFlow achieves substantial computational gain in repeated application, especially for two-dimensional LGCPs. We demonstrate the utility and robustness of the method by applying it to two distinct oral microbial biofilm images.