Search

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

In the 21st century, education has become America's most significant marker of class privilege. People with bachelor's and advanced degrees have mostly prospered, while wages for those with less education have fallen. This divergence in economic fortunes lies at the heart of our country's present economic and political discontents. Federal and state policies remain strongly biased in favor of subsidies to Americans who go to college, specifically those who acquire a 4-year degree. The nation's chronic underinvestment in work-related learning, experiences, and supports is not just unjust, it is bad economics. It squanders our most precious resource -- our workers -- and subtracts from U.S. productivity growth and competitiveness. American workers deserve better than a binary choice between an overpriced college degree program and a patchwork of public and private job training programs of uneven quality. Instead, U.S. leaders should equalize opportunity for workers, regardless of what path they choose. This means building a world-class education system that is inclusive of all options, not just college, to ensure greater upward mobility for American workers. This report is the first in a series of Campaign Blueprints detailing new ideas that can help Democrats reach across today's yawning "diploma divide" and reconnect with the working-class voters who have historically been the party's mainstay.

There is substantial variation in the returns to a college degree. One determinant is whether a worker's employment is "matched" with their education. With a novel education-industry crosswalk and panel data on 295,000 graduates, we provide the first estimates of an education-industry match premium leveraging within-person variation in earnings. We document which majors have the most and least matching, how earnings premia vary across fields and gender, and how premia evolve over time. With robust estimators, we show that workers in industries "matched" with their degree experience an average earnings premium of 7-11%, with variation by degree level and major.

The background galaxies in Euclid Early Release Observations images of the Perseus cluster make up a remarkable sample in its combination of 0.57 deg$^2$ area, 25.3 and 23.2 AB mag depth, as well as 0.1" and 0.3" angular resolutions, in optical and near-IR bands, respectively. Towards characterising the history of the Hubble sequence, we perform a preliminary morphological analysis of 2445 and 12,786 galaxies with $I_E < 21$ and $I_E < 23$, respectively. We use single-S\'ersic profiles and the sums of a S\'ersic bulge and an exponential disc to model these galaxies with SourceXtractor++ and analyse their parameters in order to assess their consistencies and biases. The fitted galaxies to $I_E < 21$ span the various Hubble types with ubiquitous bulge and disc components, and a bulge-to-total light ratio B/T taking all values from 0 to 1. The median effective radius of the single-S\'ersic profile is a biased estimate of galaxy size, intermediate between the bulge and disc effective radii, depending on B/T. The axis ratio of the single-S\'ersic profile overestimates that of the disc, increasingly so with B/T. The model impacts the photometry with -0.08 to 0.01 mag median systematic $I_E$ offsets between single-S\'ersic and bulge+disc total magnitudes, and a 0.05 to 0.15 mag dispersion, from low to high B/T. We measure a median 0.4 mag bulge-disc colour difference in $I_E - J_E$ that originates from the disc-dominated galaxies, whereas bulge-dominated galaxies have similar median colours of their components. Remarkably, we also measure redder-inside disc colour gradients, based on 5 to 10% systematic variations of disc effective radii between the optical and near-IR bands. This analysis demonstrates the usefulness and limits of single-S\'ersic profile modelling, and the power of bulge-disc decomposition for reliably characterising the morphology of lenticulars and spirals in Euclid images., Comment: Submitted to A&A, comments welcome. 27 pages, 19 figures, 6 tables (+ appendix of 6 pages, 18 figures - exemples gallery, 1 table)

Reinforcement learning (RL) has shown impressive results in sequential decision-making tasks. Meanwhile, Large Language Models (LLMs) and Vision-Language Models (VLMs) have emerged, exhibiting impressive capabilities in multimodal understanding and reasoning. These advances have led to a surge of research integrating LLMs and VLMs into RL. In this survey, we review representative works in which LLMs and VLMs are used to overcome key challenges in RL, such as lack of prior knowledge, long-horizon planning, and reward design. We present a taxonomy that categorizes these LLM/VLM-assisted RL approaches into three roles: agent, planner, and reward. We conclude by exploring open problems, including grounding, bias mitigation, improved representations, and action advice. By consolidating existing research and identifying future directions, this survey establishes a framework for integrating LLMs and VLMs into RL, advancing approaches that unify natural language and visual understanding with sequential decision-making., Comment: 9 pages, 4 figures

Dense gas in molecular clouds is an important signature of ongoing and future star formation. We identify and track dense cores in the STARFORGE simulations, following the core evolution from birth through dispersal by stellar feedback for typical Milky Way cloud conditions. Only $\sim$8% of cores host protostars, and most disperse before forming stars. The median starless and protostellar core lifetimes are $\sim 0.5-0.6$ Myr and $\sim0.8-1.1$ Myr, respectively, where the protostellar phase lasts $\sim 0.1^{+0.1}_{-0.05}$ Myr. While core evolution is stochastic, we find that virial ratios and linewidths decline in prestellar cores, coincident with turbulent decay. Collapse occurs over $\sim 0.1$ Myr, once the central density exceeds $\gtrsim 10^6$cm$^{-3}$. Starless cores, only, follow linewidth-size and mass-size relations, $\sigma \propto R^{0.3}$ and $M \propto R^1$. The core median mass, radius, and velocity dispersion scale weakly with the cloud magnetic field strength. We cluster the core properties and find that protostellar cores have $>80$% likelihood of belonging to three particular groups that are characterized by high central densities, compact radii, and lower virial parameters. Overall, core evolution appears to be universally set by the interplay of gravity and magnetized turbulence, while stellar feedback dictates protostellar core properties and sets the protostellar phase lifetime., Comment: 22 pages, 8 figures, accepted to ApJ

Decision makers may suffer from uncertainty induced by limited data. This may be mitigated by accounting for epistemic uncertainty, which is however challenging to estimate efficiently for large neural networks. To this extent we investigate Delta Variances, a family of algorithms for epistemic uncertainty quantification, that is computationally efficient and convenient to implement. It can be applied to neural networks and more general functions composed of neural networks. As an example we consider a weather simulator with a neural-network-based step function inside -- here Delta Variances empirically obtain competitive results at the cost of a single gradient computation. The approach is convenient as it requires no changes to the neural network architecture or training procedure. We discuss multiple ways to derive Delta Variances theoretically noting that special cases recover popular techniques and present a unified perspective on multiple related methods. Finally we observe that this general perspective gives rise to a natural extension and empirically show its benefit.

A 2020 physics critique report for Year 4 MSci Physics with Particle Physics & Cosmology with the University of Birmingham, School of Physics & Astronomy. Muon colliders are a proposed next-generation particle accelerator which benefit from the muon's fundamentality and relatively high mass to perform simultaneous high precision, high energy experiments. This critique reviews their physics potential and technological feasibility, then proposes a roadmap for how a 3 TeV muon collider could be built, concluding that it is experimentally possible for a muon collider to be operational in 30 years.

We extend the projected sensitivity of LDMX for sub-GeV dark matter (DM) to the case of dark photons produced through higher order electromagnetic moments. These moments arise from loop diagrams involving portal matter fields, along with the gauge fields of new symmetry groups. Due to the Lorentz structures, in particular the momentum dependence, of these additional interactions, the kinematic distributions expected at missing momentum/energy experiments vary with model in addition to dark photon mass. By considering four additional types of interactions -- magnetic and electric dipole, charge radius, and anapole moment -- we show that LDMX Phase-II is expected to probe the relic target of these additional dark photon models. We compare the analytic with the numerical methods for calculating the dark bremsstrahlung cross section, and compute the kinematic distributions for each model. The potential for model discrimination in the scenario of non-zero signal events at LDMX is discussed. We find that there is a degeneracy between the dark photon mass and model, which can be partially broken by considering both the energy and the transverse momentum of the recoil electron.

In the laser welding and additive manufacturing (AM) communities, the balling defect is primarily attributed to the action of fluid instabilities with a few authors suggesting other mechanisms. Without commenting on the validity of the fluid instability driven \textit{mechanism} of balling in AM, this work intends to present the most realistic analytical discussion of the balling defect driven purely by fluid instabilities. Synchrotron-based X-ray radiography of thin samples indicate that fluid instability growth rates and solidification can be comparable in magnitude and thus compete. Neglecting the action of fluid flows and heat transport, this work presents an analytical formalism which accounts for fluid instabilities and solidification competition, giving a continuous transition from balling to non-balling which is lacking in current literature. We adapt a Rivulet instability model from the fluid physics community to account for the stabilizing effects of the substrate which the Plateau-Rayleigh instability model does not account for, and estimate the instability growth rate. Our model predicts instability growth at higher wavelengths and shallower melt pool depths relative to width, as well as strong sensitivity to the solidification front curvature. Deviations between model predictions and our experimental results demonstrate the importance of fluid flows and heat transport in the balling process. Our experiments further demonstrate at least one mechanism by which the melt pool length and balling wavelength are not equivalent, as commonly claimed.

Microbial motion is typically analyzed by simplified models in which trajectories exhibit straight runs (perhaps with added Gaussian noise) followed by random, discrete tumbling events. We present the results of a statistical analysis of the angular dynamics for four different swimming microbes: tumbling and smooth-swimming strains of \textit{Bacillus subtilis} and two Eukaryotic algae, \textit{Tetraselmis suecica} and \textit{Euglena gracilis}. We show that the angular statistics closely resemble a Voigt profile, the convolution of a Gaussian (L\'{e}vy index $\alpha=2$) and Lorentzian (L\'{e}vy index $\alpha=1$) distribution. This distribution is ubiquitous for all four microbes. Rather than modeling tumbling as a discrete process, we model tumbling dynamics as a continuous process: L\'{e}vy flights in the orientational dynamics using a Lorentzian noise model. This model is analytically solvable. Each individual microbe trajectory has both stochastic behavior (noise) and varying deterministic behavior, such as helices of different sizes and frequencies and circular arcs with different radii. We model the distribution of different deterministic behavior via an ensemble theory. The deterministic behavior (e.g., circular arcs) comes from physical observations of the swimming behavior and explains many of the qualitative features in the data that cannot be explained by a pure noise model. From this theory, we estimate the strength of Lorentzian noise, the physical rotational diffusion constant, and some relevant parameters relating to the distributions of deterministic behavior. This analysis shows that in some cases Gaussian noise is not the dominant process responsible for the angular statistics following a Voigt profile., Comment: 17 pages, 16 figures

Ram pressure stripping is a well-known environmental quenching mechanism that removes gas from galaxies infalling into groups and clusters. In some extreme examples of ram pressure stripping, galaxies with extended gas tails show evidence of enhanced star formation prior to quenching. In this work we use a sample of 5277 local satellite galaxies in which a stripped tail of gas has not necessarily been observed, to quantify the strength of ram pressure-enhanced star formation and compare these results to a control sample of 8360 field galaxies. We use u-band imaging from the Ultraviolet-Near Infrared Northern Survey (UNIONS) as a star formation tracer and several metrics to quantify star formation asymmetry. We compare these results to environmental properties of the galaxy, such as their time since infall and host halo mass, to constrain the degree of ram pressure enhanced star formation as a function of environment. We find no significant differences between the satellite and the field samples. We further restrict our sample to galaxies which we most expect to be experiencing significant ram pressure but find no strong evidence of these galaxies having systematically enhanced star formation. Finally, we investigate the properties of the most asymmetric galaxies in our sample and again find no strong evidence of ram pressure-induced star formation enhancement. We conclude that any star formation enhancement must be small for infalling galaxies, suggesting that this effect is either uncommon or short-lived., Comment: 19 pages, 10 figures, accepted by ApJ

How should one judge whether a given large language model (LLM) can reliably perform economic reasoning? Most existing LLM benchmarks focus on specific applications and fail to present the model with a rich variety of economic tasks. A notable exception is Raman et al. [2024], who offer an approach for comprehensively benchmarking strategic decision-making; however, this approach fails to address the non-strategic settings prevalent in microeconomics, such as supply-and-demand analysis. We address this gap by taxonomizing microeconomic reasoning into $58$ distinct elements, focusing on the logic of supply and demand, each grounded in up to $10$ distinct domains, $5$ perspectives, and $3$ types. The generation of benchmark data across this combinatorial space is powered by a novel LLM-assisted data generation protocol that we dub auto-STEER, which generates a set of questions by adapting handwritten templates to target new domains and perspectives. Because it offers an automated way of generating fresh questions, auto-STEER mitigates the risk that LLMs will be trained to over-fit evaluation benchmarks; we thus hope that it will serve as a useful tool both for evaluating and fine-tuning models for years to come. We demonstrate the usefulness of our benchmark via a case study on $27$ LLMs, ranging from small open-source models to the current state of the art. We examined each model's ability to solve microeconomic problems across our whole taxonomy and present the results across a range of prompting strategies and scoring metrics., Comment: 18 pages, 11 figures

We describe a framework and tool specification that represents a step towards cybersecurity testing and monitoring of IoT ecosystems. We begin with challenges from a previous paper and discuss an integrated approach and tools to enable testing and monitoring to address these challenges. We also describe exemplary use cases of IoT ecosystems and propose approaches to address the challenges using the framework and tools. The current status of this work is that the specification and conceptualisation is complete, use cases are understood with clear challenges and implementation / extension of the tools and framework is underway with tools at different stages of development. Several key observations have been made throughout this work, as follows. 1) Tools may be used in multiple different combinations, and ad-hoc use is also encouraged, where one tool may provide clues and other tools executed to undertake further investigations based on initial results. 2) Automated execution of tool chains is supported by workflows. 3) support for immutable storage of audit records of tests and results is an important requirement. 4) Indicators (observations or measurements representing information of relevance for assessment of cyber security) are a key mechanism for intercommunication between one tool and another, or with the operator. 5) Mapping this work to established security development lifecycles is a useful means of determining applicability and utility of the tools and framework. 6) There is a key interplay between devices and systems. 7) Anomaly detection in multiple forms is a key means of runtime monitoring. 8) Considerable investigation is needed related to the specifics of each device / system as an item of further work., Comment: Preprint of Paper submitted to Springer Nature Computer Science

Disorder remains a major obstacle to realizing topological Majorana zero modes (MZMs) in superconductor-semiconductor nanowires, and we show how deep learning can be used to recover topological MZMs mitigating disorder even when the pre-mitigation situation manifests no apparent topology. The disorder potential, as well as the scattering invariant ($T_V$) normally used to classify a device as topologically non-trivial are not directly measurable experimentally. Additionally, the conventional signatures of MZMs have proved insufficient due to their being accidentally replicated by disorder-induced trivial states. Recent advances in machine learning provide a novel method to solve these problems, allowing the underlying topology, suppressed by disorder, to be recovered using effective mitigation procedures. In this work, we leverage a vision transformer neural network trained on conductance measurements along with a CMA-ES optimization framework to dynamically tune gate voltages mitigating disorder effects. Unlike prior efforts that relied on indirect cost functions, our method directly optimizes $T_V$ alongside additional local density of states-based topological indicators. Using a lightweight neural network variant, we demonstrate that even highly disordered nanowires initially lacking any topologically non-trivial regions can be transformed into robust topological devices., Comment: 14 figures, 10 pages

This work demonstrates that diffusion models can achieve font-controllable multilingual text rendering using just raw images without font label annotations. Visual text rendering remains a significant challenge. While recent methods condition diffusion on glyphs, it is impossible to retrieve exact font annotations from large-scale, real-world datasets, which prevents user-specified font control. To address this, we propose a data-driven solution that integrates the conditional diffusion model with a text segmentation model, utilizing segmentation masks to capture and represent fonts in pixel space in a self-supervised manner, thereby eliminating the need for any ground-truth labels and enabling users to customize text rendering with any multilingual font of their choice. The experiment provides a proof of concept of our algorithm in zero-shot text and font editing across diverse fonts and languages, providing valuable insights for the community and industry toward achieving generalized visual text rendering., Comment: This is preliminary work and code will be released at github.com/bowen-upenn/ControlText

This paper presents a hybrid approach that integrates Large Language Models (LLMs) with a multi-scenario Stochastic Unit Commitment (SUC) framework, focusing on both efficiency and reliability under high wind generation uncertainties. Numerical experiments on small-to-medium-sized test systems show that while the traditional SUC approach yields a total cost of 99.05 million USD with 3.04 GWh of load curtailment, the LLM-assisted SUC (LLM-SUC) reduces costs to 98.87 million USD and lowers load curtailment to 2.32 GWh, an improvement of nearly 24%. Both methods maintain zero wind curtailment, confirming robust renewable integration. By employing an LLM agent that helps balance the energy system more effectively, the proposed framework enhances demand fulfillment at reduced costs, illustrating the potential of AI to inform generator commitments in uncertain operating conditions. Further gains may be realized by refining prompt design, incorporating historical operational data, and extending this approach to higher-dimensional uncertainties and energy storage systems, ultimately fostering greater resilience, efficiency, and adaptability in next-generation power system operations.

We explore the occurrence of point configurations within non-meager (second category) Baire sets. A celebrated result of Steinhaus asserts that $A+B$ and $A-B$ contain an interval whenever $A$ and $B$ are sets of positive Lebesgue measure in $\R^n$ for $n\geq 1$. A topological analogue attributed to Piccard asserts that both $AB$ and $AB^{-1}$ contain an interval when $A,B$ are non-meager (second category) Baire sets in a topological group. We explore generalizations of Piccard's result to more complex point configurations and more abstract spaces. In the Euclidean setting, we show that if $A\subset \R^d$ is a non-meager Baire set and $F=\{x_n\}_{n\in\N}$ is a bounded sequence, then there is an interval of scalings $t$ for which $tF+z\subset A$ for some $z\in \R^d$. That is, the set $$\Delta_F(A)=\{t\in\R: \exists z\text{ such that }tF+z\subset A\}$$ has nonempty interior. More generally, if $V$ is a topological vector space and $F=\{x_n\}_{n\in\N}\subset V$ is a bounded sequence, we show that if $A\subset V$ is non-meager and Baire, then $\Delta_F(A)$ has nonempty interior. The notion of boundedness in this context is described below. Note that the sequence $F$ can be countably infinite, which distinguishes this result from its measure-theoretic analogue. In the context of the topological version of Erd\H{o}s' similarity conjecture, we show that bounded countable sets are universal in non-meager Baire sets.

The Early Release Observations (ERO) from Euclid have detected several new galaxy-galaxy strong gravitational lenses, with the all-sky survey expected to find 170,000 new systems, greatly enhancing studies of dark matter, dark energy, and constraints on the cosmological parameters. As a first step, visual inspection of all galaxies in one of the ERO fields (Perseus) was carried out to identify candidate strong lensing systems and compared to the predictions from Convolutional Neural Networks (CNNs). However, the entire ERO data set is too large for expert visual inspection. In this paper, we therefore extend the CNN analysis to the whole ERO data set, using different CNN architectures and methodologies. Using five CNN architectures, we identified 8,469 strong gravitational lens candidates from IE-band cutouts of 13 Euclid ERO fields, narrowing them to 97 through visual inspection, including 14 grade A and 31 grade B candidates. We present the spectroscopic confirmation of a strong gravitational lensing candidate, EUCLJ081705.61+702348.8. The foreground lensing galaxy, an early-type system at redshift z = 0.335, and the background source, a star-forming galaxy at redshift z = 1.475 with [O II] emission, are both identified. Lens modeling using the Euclid strong lens modeling pipeline reveals two distinct arcs in a lensing configuration, with an Einstein radius of 1.18 \pm 0.03 arcseconds, confirming the lensing nature of the system. These findings highlight the importance of a broad CNN search to efficiently reduce candidates, followed by visual inspection to eliminate false positives and achieve a high-purity sample of strong lenses in Euclid.

The emergence of generative AI, particularly large language models (LLMs), has opened the door for student-centered and active learning methods like project-based learning (PBL). However, PBL poses practical implementation challenges for educators around project design and management, assessment, and balancing student guidance with student autonomy. The following research documents a co-design process with interdisciplinary K-12 teachers to explore and address the current PBL challenges they face. Through teacher-driven interviews, collaborative workshops, and iterative design of wireframes, we gathered evidence for ways LLMs can support teachers in implementing high-quality PBL pedagogy by automating routine tasks and enhancing personalized learning. Teachers in the study advocated for supporting their professional growth and augmenting their current roles without replacing them. They also identified affordances and challenges around classroom integration, including resource requirements and constraints, ethical concerns, and potential immediate and long-term impacts. Drawing on these, we propose design guidelines for future deployment of LLM tools in PBL., Comment: 25 pages

A previous case study measured performance vs source-code complexity across multiple languages. The case study identified Chapel and HPX provide similar performance and code complexity. This paper is the result of initial steps toward closing the source-code complexity gap between Chapel and HPX by using a source-to-source compiler. The investigation assesses the single-machine performance of both Chapel and Chplx applications across Arm and x86.

We present Cryoscope -- a new 50 sq. deg field-of-view, 1.2 m aperture, K-dark survey telescope to be located at Dome C, Antarctica. Cryoscope has an innovative optical-thermal design wherein the entire telescope is cryogenically cooled. Cryoscope also explores new detector technology to cost-effectively tile the full focal plane. Leveraging the dark Antarctic sky and minimizing telescope thermal emission, Cryoscope achieves unprecedented deep, wide, fast and red observations, matching and exceeding volumetric survey speeds from the Ultraviolet Explorer, Vera Rubin Observatory, and Nancy Grace Roman Space Telescope. By providing coverage beyond wavelengths of 2 $\mu$m, we aim to create the most comprehensive dynamic movie of the most obscured reaches of the Universe. Cryoscope will be a dedicated discovery engine for electromagnetic emission from coalescing compact binaries, Earth-like exoplanets orbiting cold stars, and multiple facets of time-domain, stellar and solar system science. In this paper, we describe the scientific drivers and technical innovations for this new discovery engine operating in the K-dark passband, why we choose to deploy it in Antarctica, and the status of a fifth-scale prototype designed as a Pathfinder to retire technological risks prior to full-scale implementation., Comment: 36 pages, 20 figures, 4 tables; submitted to PASP on 2025-02-09

In this note, we study a lattice point counting problem for spheres in Heisenberg groups, incorporating both the non-isotropic dilation structure and the non-commutative group law. More specifically, we establish an upper bound for the average number of lattice points in a $\delta$-neighborhood of a Kor\'anyi sphere of large radius, where the average considered is over Heisenberg group translations of the sphere. This is in contrast with previous works, which either count lattice points on dilates of a fixed sphere (see \cites{GNT, Gath2}) or consider averages over Euclidean translations of the sphere (see \cites{CT}). We observe that incorporating the Heisenberg group structure allows us to circumvent the degeneracy arising from the vanishing of the Gaussian curvature at the poles of the Kor\'anyi sphere. In fact, in lower dimensions (the first and second Heisenberg group), our method establishes an upper bound for this number which gives a logarithmic improvement over the bound implied by the previously known results. Even for the higher dimensional Heisenberg groups, we recover the bounds implied by the main result of \cite{GNT} using a completely different approach of generalized Radon transforms. Further, we obtain upper bounds for the average number of lattice points near more general spheres described with respect to radial, Heisenberg homogeneous norms as considered in \cite{GNT}., Comment: 25 pages. Comments are welcome!

We introduce and analyse a new, extremely simple, randomised sorting algorithm: - choose a pair of indices $\{i, j\}$ according to some distribution $q$; - sort the elements in positions $i$ and $j$ of the array in ascending order. Choosing $q_{\{i,j\}} \propto 1/|j - i|$ yields an order-$n (\log n)^2$ sorting time. We call it the harmonic sorter. The sorter trivially parallelises in the asynchronous setting, yielding a linear speed-up. We also exhibit a low-communication, synchronous version with a linear speed-up. We compare and contrast this algorithm with other sorters, and discuss some of its benefits, particularly its robustness and amenability to parallelisation and distributed computing., Comment: 21 pages: 18 body + 3 appendix

Electrocardiogram (ECG) analysis has emerged as a promising tool for identifying physiological changes associated with neuropsychiatric conditions. The relationship between cardiovascular health and neuropsychiatric disorders suggests that ECG abnormalities could serve as valuable biomarkers for more efficient detection, therapy monitoring, and risk stratification. However, the potential of the ECG to accurately distinguish neuropsychiatric conditions, particularly among diverse patient populations, remains underexplored. This study utilized ECG markers and basic demographic data to predict neuropsychiatric conditions using machine learning models, with targets defined through ICD-10 codes. Both internal and external validation were performed using the MIMIC-IV and ECG-View datasets respectively. Performance was assessed using AUROC scores. To enhance model interpretability, Shapley values were applied to provide insights into the contributions of individual ECG features to the predictions. Significant predictive performance was observed for conditions within the neurological and psychiatric groups. For the neurological group, Alzheimer's disease (G30) achieved an internal AUROC of 0.813 (0.812-0.814) and an external AUROC of 0.868 (0.867-0.868). In the psychiatric group, unspecified dementia (F03) showed an internal AUROC of 0.849 (0.848-0.849) and an external AUROC of 0.862 (0.861-0.863). Discriminative features align with known ECG markers but also provide hints on potentially new markers. ECG offers significant promise for diagnosing and monitoring neuropsychiatric conditions, with robust predictive performance across internal and external cohorts. Future work should focus on addressing potential confounders, such as therapy-related cardiotoxicity, and expanding the scope of ECG applications, including personalized care and early intervention strategies., Comment: 9 pages, 2 figures, source code under https://github.com/AI4HealthUOL/CardioDiag

Normative mapping is a framework used to map population-level features of health-related variables. It is widely used in neuroscience research, but the literature lacks established protocols in modalities that do not support healthy control measurements, such as intracranial EEG (icEEG). An icEEG normative map would allow researchers to learn about population-level brain activity and enable comparison of individual data against these norms to identify abnormalities. Currently, no standardised guide exists for transforming clinical data into a normative, regional icEEG map. Papers often cite different software and numerous articles to summarise the lengthy method, making it laborious for other researchers to understand or apply the process. Our protocol seeks to remedy this gap by providing a dataflow guide and key decision points that summarise existing methods. This protocol is used heavily in published works from our own lab (twelve peer-reviewed journal publications). Briefly, we take as input, icEEG recordings and neuroimaging data from people with epilepsy who are undergoing evaluation for resective surgery. As final outputs, we obtain a normative icEEG map, comprising signal properties localised to brain regions. Optionally, we can also process new subjects through the same pipeline and obtain their z-scores (or centiles) in each brain region, for abnormality detection and localisation. To date, a single, cohesive, dataflow pipeline for generating normative icEEG maps, along with abnormality mapping, has not been created. We envisage that this dataflow guide will not only increase understanding and application of normative mapping methods, but will also improve the consistency and quality of studies in the field.

Dark Scattering (DS) is an interacting dark energy model characterised by pure momentum exchange between dark energy and dark matter. It is phenomenologically interesting because it is unconstrained by CMB data and can alleviate the $S_8$ tension. We derive constraints on cosmological and DS parameters using three two-point correlation functions (3$\times$2pt) from the Dark Energy Survey third year data release (DES Y3). We then add information from the multipoles of the galaxy power spectrum combined with Baryonic Acoustic Oscillation (BAO) measurements using the twelfth data release of the Baryon Oscillation Spectroscopic Survey (BOSS DR12) and external BAO measurements. We compare results from the direct combination of the probes with the joint posterior distribution calculated with a normalising flow approach. Additionally, we run a CMB analysis with the Planck Public Release 4 (PR4) for comparison of the cosmological constraints. Overall, we find that the combination of probes allows minimising the projection effects and improves constraints without the need to include CMB information. It brings the marginalised posterior maxima closer to the corresponding best-fit values and weakens the sensitivity to the priors of the spectroscopic modelling nuisance parameters. These findings are highly relevant in light of forthcoming data of surveys like DESI, Euclid, and Rubin.

Self-play has powered breakthroughs in two-player and multi-player games. Here we show that self-play is a surprisingly effective strategy in another domain. We show that robust and naturalistic driving emerges entirely from self-play in simulation at unprecedented scale -- 1.6~billion~km of driving. This is enabled by Gigaflow, a batched simulator that can synthesize and train on 42 years of subjective driving experience per hour on a single 8-GPU node. The resulting policy achieves state-of-the-art performance on three independent autonomous driving benchmarks. The policy outperforms the prior state of the art when tested on recorded real-world scenarios, amidst human drivers, without ever seeing human data during training. The policy is realistic when assessed against human references and achieves unprecedented robustness, averaging 17.5 years of continuous driving between incidents in simulation.

Bayesian optimization (BO) has emerged as a powerful tool for guiding experimental design and decision-making in various scientific fields, including materials science, chemistry, and biology. However, despite its growing popularity, the complexity of existing BO libraries and the steep learning curve associated with them can deter researchers who are not well-versed in machine learning or programming. To address this barrier, we introduce Honegumi, a user-friendly, interactive tool designed to simplify the process of creating advanced Bayesian optimization scripts. Honegumi offers a dynamic selection grid that allows users to configure key parameters of their optimization tasks, generating ready-to-use, unit-tested Python scripts tailored to their specific needs. Accompanying the interface is a comprehensive suite of tutorials that provide both conceptual and practical guidance, bridging the gap between theoretical understanding and practical implementation. Built on top of the Ax platform, Honegumi leverages the power of existing state-of-the-art libraries while restructuring the user experience to make advanced BO techniques more accessible to experimental researchers. By lowering the barrier to entry and providing educational resources, Honegumi aims to accelerate the adoption of advanced Bayesian optimization methods across various domains., Comment: 7 pages, 3 figures, 1 table

Visible watermarks pose significant challenges for image restoration techniques, especially when the target background is unknown. Toward this end, we present MorphoMod, a novel method for automated visible watermark removal that operates in a blind setting -- without requiring target images. Unlike existing methods, MorphoMod effectively removes opaque and transparent watermarks while preserving semantic content, making it well-suited for real-world applications. Evaluations on benchmark datasets, including the Colored Large-scale Watermark Dataset (CLWD), LOGO-series, and the newly introduced Alpha1 datasets, demonstrate that MorphoMod achieves up to a 50.8% improvement in watermark removal effectiveness compared to state-of-the-art methods. Ablation studies highlight the impact of prompts used for inpainting, pre-removal filling strategies, and inpainting model performance on watermark removal. Additionally, a case study on steganographic disorientation reveals broader applications for watermark removal in disrupting high-level hidden messages. MorphoMod offers a robust, adaptable solution for watermark removal and opens avenues for further advancements in image restoration and adversarial manipulation.

In this chapter we begin with a review of Titan's fluvial and lacustrine landscapes as observed with Cassini remote sensing data, and what the many discoveries have revealed about Titan's surface materials and climate. Yet Cassini remote sensing data are coarse, topographic data are largely lacking, and the absence of in situ field measurements means we have little understanding of what the surface is composed of. At present, our knowledge of Titan's hydrology is comparable to that of Mars in the 1970's during the Viking era. Fortunately, the coming decades promise many new and exciting discoveries that can be achieved through Earth-based experiments, numerical modeling, and a continued commitment to the exploration of Titan by future missions, including both Dragonfly and orbiting assets. We therefore close the chapter with a discussion about what can be done with the current Cassini data and how new data, from both Dragonfly and a potential future orbiter, would allow us to leverage Titan to help solve some of the largest problems both here on Earth and on hydrologic planets and exoplanets more generally., Comment: 26 pages, 7 figures

We present a novel prompt design for Large Language Models (LLMs) tailored to Asynchronous Time Series. Unlike regular time series, which assume values at evenly spaced time points, asynchronous time series consist of timestamped events occurring at irregular intervals, each described in natural language. Our approach effectively utilizes the rich natural language of event descriptions, allowing LLMs to benefit from their broad world knowledge for reasoning across different domains and tasks. This allows us to extend the scope of asynchronous time series analysis beyond forecasting to include tasks like anomaly detection and data imputation. We further introduce Stochastic Soft Prompting, a novel prompt-tuning mechanism that significantly improves model performance, outperforming existing fine-tuning methods such as QLoRA. Through extensive experiments on real world datasets, we demonstrate that our approach achieves state-of-the-art performance across different tasks and datasets.

Pre-trained transformer models with extended context windows are notoriously expensive to run at scale, often limiting real-world deployment due to their high computational and memory requirements. In this paper, we introduce Hamming Attention Distillation (HAD), a novel framework that binarizes keys and queries in the attention mechanism to achieve significant efficiency gains. By converting keys and queries into {-1, +1} vectors and replacing dot-product operations with efficient Hamming distance computations, our method drastically reduces computational overhead. Additionally, we incorporate attention matrix sparsification to prune low-impact activations, which further reduces the cost of processing long-context sequences. \par Despite these aggressive compression strategies, our distilled approach preserves a high degree of representational power, leading to substantially improved accuracy compared to prior transformer binarization methods. We evaluate HAD on a range of tasks and models, including the GLUE benchmark, ImageNet, and QuALITY, demonstrating state-of-the-art performance among binarized Transformers while drastically reducing the computational costs of long-context inference. \par We implement HAD in custom hardware simulations, demonstrating superior performance characteristics compared to a custom hardware implementation of standard attention. HAD achieves just $\mathbf{1.78}\%$ performance losses on GLUE compared to $9.08\%$ in state-of-the-art binarization work, and $\mathbf{2.5}\%$ performance losses on ImageNet compared to $12.14\%$, all while targeting custom hardware with a $\mathbf{79}\%$ area reduction and $\mathbf{87}\%$ power reduction compared to its standard attention counterpart.

The most sensitive direct neutrino mass searches today are based on measurement of the endpoint of the beta spectrum of tritium to infer limits on the mass of the unobserved recoiling neutrino. To avoid the smearing associated with the distribution of molecular final states in the T-He molecule, the next generation of these experiments will need to employ atomic (T) rather than molecular (T$_{2}$) tritium sources. Following production, atomic T can be trapped in gravitational and / or magnetic bottles for beta spectrum experiments, if and only if it can first be cooled to millikelvin temperatures. Accomplishing this cooling presents substantial technological challenges. The Project 8 collaboration is developing a technique based on magnetic evaporative cooling along a beamline (MECB) for the purpose of cooling T to feed a magneto-gravitational trap that also serves as a cyclotron radiation emission spectroscope. Initial tests of the approach are planned in a pathfinder apparatus using atomic Li. This paper presents a method for analyzing the dynamics of the MECB technique, and applies these calculations to the design of systems for cooling and slowing of atomic Li and T. A scheme is outlined that could provide a current of T at the millikelvin temperatures required for the Project 8 neutrino mass search.