Search

We use student-level data on elementary special education identification from Washington state to explore student identification rates in the months immediately after the onset of the COVID-19 pandemic and over 2 subsequent years. Special education identification rates dropped dramatically in March 2020 through the end of the 2019-20 school year and remained below historical norms through 2020-21 before returning to pre-pandemic levels early in 2021-22. The magnitude of these effects implies that over 8,000 fewer elementary students were identified for special education services during the pandemic in Washington than would have been expected based on prior trends.

Neural network verification is a new and rapidly developing field of research. So far, the main priority has been establishing efficient verification algorithms and tools, while proper support from the programming language perspective has been considered secondary or unimportant. Yet, there is mounting evidence that insights from the programming language community may make a difference in the future development of this domain. In this paper, we formulate neural network verification challenges as programming language challenges and suggest possible future solutions., Comment: Accepted at ESOP 2025, European Symposium on Programming Languages

We contribute to the lively debate in current scholarship on the Leibnizian calculus. In a recent text, Arthur and Rabouin argue that non-Archimedean continua are incompatible with Leibniz's concepts of number, quantity and magnitude. They allege that Leibniz viewed infinitesimals as contradictory, and claim to deduce such a conclusion from an analysis of the Leibnizian definition of quantity. However, their argument is marred by numerous errors, deliberate omissions, and misrepresentations, stemming in a number of cases from flawed analyses in their earlier publications. We defend the thesis, traceable to the classic study by Henk Bos, that Leibniz used genuine infinitesimals, which he viewed as fictional mathematical entities (and not merely shorthand for talk about more ordinary quantities) on par with negatives and imaginaries., Comment: 38 pages; to appear in Ghanita Bharati

Modern Language Models (LMs) owe much of their success to masked causal attention, the backbone of Generative Pre-Trained Transformer (GPT) models. Although GPTs can process the entire user prompt at once, the causal masking is applied to all input tokens step-by-step, mimicking the generation process. This imposes an unnecessary constraint during the initial "prefill" phase when the model processes the input prompt and generates the internal representations before producing any output tokens. In this work, attention is masked based on the known block structure at the prefill phase, followed by the conventional token-by-token autoregressive process after that. For example, in a typical chat prompt, the system prompt is treated as one block, and the user prompt as the next one. Each of these is treated as a unit for the purpose of masking, such that the first tokens in each block can access the subsequent tokens in a non-causal manner. Then, the model answer is generated in the conventional causal manner. This Segment-by-Segment scheme entails no additional computational overhead. When integrating it into models such as Llama and Qwen, state-of-the-art performance is consistently achieved.

The success of Transformer-based Language Models (LMs) stems from their attention mechanism. While this mechanism has been extensively studied in explainability research, particularly through the attention values obtained during the forward pass of LMs, the backward pass of attention has been largely overlooked. In this work, we study the mathematics of the backward pass of attention, revealing that it implicitly calculates an attention matrix we refer to as "Reversed Attention". We examine the properties of Reversed Attention and demonstrate its ability to elucidate the models' behavior and edit dynamics. In an experimental setup, we showcase the ability of Reversed Attention to directly alter the forward pass of attention, without modifying the model's weights, using a novel method called "attention patching". In addition to enhancing the comprehension of how LM configure attention layers during backpropagation, Reversed Attention maps contribute to a more interpretable backward pass.

As LLM-based applications reach millions of customers, ensuring their scalability and continuous quality improvement is critical for success. However, the current workflows for developing, maintaining, and operating (DevOps) these applications are predominantly manual, slow, and based on trial-and-error. With this paper we introduce the Generative AI Toolkit, which automates essential workflows over the whole life cycle of LLM-based applications. The toolkit helps to configure, test, continuously monitor and optimize Generative AI applications such as agents, thus significantly improving quality while shortening release cycles. We showcase the effectiveness of our toolkit on representative use cases, share best practices, and outline future enhancements. Since we are convinced that our Generative AI Toolkit is helpful for other teams, we are open sourcing it on and hope that others will use, forward, adapt and improve, Comment: 16 pages, 6 figures. For source code see https://github.com/awslabs/generative-ai-toolkit

Phishing attacks on enterprise employees present one of the most costly and potent threats to organizations. We explore an understudied facet of enterprise phishing attacks: the email relay infrastructure behind successfully delivered phishing emails. We draw on a dataset spanning one year across thousands of enterprises, billions of emails, and over 800,000 delivered phishing attacks. Our work sheds light on the network origins of phishing emails received by real-world enterprises, differences in email traffic we observe from networks sending phishing emails, and how these characteristics change over time. Surprisingly, we find that over one-third of the phishing email in our dataset originates from highly reputable networks, including Amazon and Microsoft. Their total volume of phishing email is consistently high across multiple months in our dataset, even though the overwhelming majority of email sent by these networks is benign. In contrast, we observe that a large portion of phishing emails originate from networks where the vast majority of emails they send are phishing, but their email traffic is not consistent over time. Taken together, our results explain why no singular defense strategy, such as static blocklists (which are commonly used in email security filters deployed by organizations in our dataset), is effective at blocking enterprise phishing. Based on our offline analysis, we partnered with a large email security company to deploy a classifier that uses dynamically updated network-based features. In a production environment over a period of 4.5 months, our new detector was able to identify 3-5% more enterprise email attacks that were previously undetected by the company's existing classifiers., Comment: To appear in the proceedings of the Passive and Active Network Measurement (PAM 2025)

We present efficient algorithms for obtaining the Hamiltonian in Lightcone Conformal Truncation (LCT) for a 2d scalar field with a generic potential. We apply this method to the sine-Gordon and sinh-Gordon models in $1+1d$, and find precise agreement with integrability results when the scaling dimension $\Delta$ of the deforming cosine/cosinh potential is in the range $ \Delta \leq 1$. The agreement provides additional evidence for a recent conjecture for how to compute the effective lightcone Hamiltonian in this class of models. In addition, to high precision, we provide the first direct confirmation for the conjectured self-duality of the sinh-Gordon model ($\Delta < 0)$, which relates $\Delta \leftrightarrow 4/\Delta$. As the dimension approaches the upper limit $\Delta=1$ from below, we show analytically that the Hamiltonian matrix elements exactly reproduce those of a free Majorana fermion, demonstrating how bosonization is manifested in the LCT basis. We comment on the possible extension of the approach to $\Delta > 1$., Comment: 30 pages, 11 figures

The Common Workflow Language (CWL) is a widely adopted language for defining and sharing computational workflows. It is designed to be independent of the execution engine on which workflows are executed. In this paper, we describe our experiences integrating CWL with Parsl, a Python-based parallel programming library designed to manage execution of workflows across diverse computing environments. We propose a new method that converts CWL CommandLineTool definitions into Parsl apps, enabling Parsl scripts to easily import and use tools represented in CWL. We describe a Parsl runner that is capable of executing a CWL CommandLineTool directly. We also describe a proof-of-concept extension to support inline Python in a CWL workflow definition, enabling seamless use in the Python ecosystem of Parsl. We demonstrate the benefits of this integration by presenting example CWL CommandLineTool definitions that show how they can be used in Parsl, and comparing performance of executing an image processing workflow using the Parsl integration and other CWL runners., Comment: 8 pages, 3 figures, IEEE/ACM Supercomputing Conference (SC24)

The very long-term evolution of the hierarchical restricted three-body problem with a slightly aligned precessing quadrupole potential is investigated analytically and solved for both rotating and librating Kozai-Lidov cycles (KLCs) with high eccentricities. We describe the finding of a striking similarity between librating and rotating KLCs for some range of precession rates. We show that the main effect occurs in both categories when the KLC frequency is equal to the precession rate of the perturbing potential. We solve the resonant dynamics analytically and show that it is equivalent to a simple pendulum model allowing us to map the strikingly rich structures that arise for precession rates similar to the Kozai-Lidov timescale (ratio of a few) and explain the similarity and when it vanishes. Additionally, we show that the regular KLCs at high eccentricities can also be described using a simple pendulum., Comment: 6 pages, 5 figures

Automated decision-making is a fundamental topic that spans multiple sub-disciplines in AI: reinforcement learning (RL), AI planning (AP), foundation models, and operations research, among others. Despite recent efforts to ``bridge the gaps'' between these communities, there remain many insights that have not yet transcended the boundaries. Our goal in this paper is to provide a brief and non-exhaustive primer on ideas well-known in AP, but less so in other sub-disciplines. We do so by introducing the classical AP problem and representation, and extensions that handle uncertainty and time through the Markov Decision Process formalism. Next, we survey state-of-the-art techniques and ideas for solving AP problems, focusing on their ability to exploit problem structure. Lastly, we cover subfields within AP for learning structure from unstructured inputs and learning to generalise to unseen scenarios and situations.

We analyzed the 7.92$\times 10^{11}$ cosmic-ray-induced muon events collected by the IceCube Neutrino Observatory from May 13, 2011, when the fully constructed experiment started to take data, to May 12, 2023. This dataset provides an up-to-date cosmic-ray arrival direction distribution in the Southern Hemisphere with unprecedented statistical accuracy covering more than a full period length of a solar cycle. Improvements in Monte Carlo event simulation and better handling of year-to-year differences in data processing significantly reduce systematic uncertainties below the level of statistical fluctuations compared to the previously published results. We confirm the observation of a change in the angular structure of the cosmic-ray anisotropy between 10 TeV and 1 PeV, more specifically in the 100-300 TeV energy range.

Robust inference of galaxy stellar masses from photometry is crucial for constraints on galaxy assembly across cosmic time. Here, we test a commonly-used Spectral Energy Distribution (SED) fitting code, using simulated galaxies from the SPHINX20 cosmological radiation hydrodynamics simulation, with JWST NIRCam photometry forward-modelled with radiative transfer. Fitting the synthetic photometry with various star formation history models, we show that recovered stellar masses are, encouragingly, generally robust to within a factor of ~3 for galaxies in the range M*~10^7-10^9M_sol at z=5-10. These results are in stark contrast to recent work claiming that stellar masses can be underestimated by as much as an order of magnitude in these mass and redshift ranges. However, while >90% of masses are recovered to within 0.5dex, there are notable systematic trends, with stellar masses typically overestimated for low-mass galaxies (M*<~10^8M_sol) and slightly underestimated for high-mass galaxies (M*>~10^9M_sol). We demonstrate that these trends arise due to the SED fitting code poorly modelling the impact of strong emission lines on broadband photometry. These systematic trends, which exist for all star formation history parametrisations tested, have a tilting effect on the inferred stellar mass function, with number densities of massive galaxies underestimated (particularly at the lowest redshifts studied) and number densities of lower-mass galaxies typically overestimated. Overall, this work suggests that we should be optimistic about our ability to infer the masses of high-z galaxies observed with JWST (notwithstanding contamination from AGN) but careful when modelling the impact of strong emission lines on broadband photometry., Comment: 18 pages, 7 figures. Accepted for publication in ApJL

Estimating the probability of failure is a critical step in developing safety-critical autonomous systems. Direct estimation methods such as Monte Carlo sampling are often impractical due to the rarity of failures in these systems. Existing importance sampling approaches do not scale to sequential decision-making systems with large state spaces and long horizons. We propose an adaptive importance sampling algorithm to address these limitations. Our method minimizes the forward Kullback-Leibler divergence between a state-dependent proposal distribution and a relaxed form of the optimal importance sampling distribution. Our method uses Markov score ascent methods to estimate this objective. We evaluate our approach on four sequential systems and show that it provides more accurate failure probability estimates than baseline Monte Carlo and importance sampling techniques. This work is open sourced., Comment: Submitted to L4DC 2025

Combinatorial mechanical metamaterials are made of anisotropic, flexible blocks, such that multiple metamaterials may be constructed using a single block type, and the system's response depends on the frustration (or its absence) due to the mutual orientations of the blocks within the lattice. Specifically, any minimal loop of blocks that may not simultaneously deform in their softest mode defines a mechanical defect at the vertex (in two dimensions) or edge (in three dimensions) that the loop encircles. Defects stiffen the metamaterial, and allow to design the spatial patterns of stress and deformation as the system is externally loaded. We study the ability to place defects at arbitrary positions in metamaterials made of a family of block types that we recently introduced for the square, honeycomb, and cubic lattices. Alongside blocks for which we show that any defect configuration is possible, we identify situations in which not all sets are realizable as defects. One of the restrictions is that in three dimensions, defected edges form closed curves. Even in cases when not all geometries of defect lines are possible, we show how to produce defect lines of arbitrary knottedness., Comment: See accompanying paper: Breaking Mechanical Holography in Combinatorial Metamaterials arXiv:2411.15760

Due to the inherent lack of transparency in deep neural networks, it is challenging for deep reinforcement learning (DRL) agents to gain trust and acceptance from users, especially in safety-critical applications such as medical diagnosis and military operations. Existing methods for explaining an agent's decision either require to retrain the agent using models that support explanation generation or rely on perturbation-based techniques to reveal the significance of different input features in the decision making process. However, retraining the agent may compromise its integrity and performance, while perturbation-based methods have limited performance and lack knowledge accumulation or learning capabilities. Moreover, since each perturbation is performed independently, the joint state of the perturbed inputs may not be physically meaningful. To address these challenges, we introduce $\textbf{VisionMask}$, a standalone explanation model trained end-to-end to identify the most critical regions in the agent's visual input that can explain its actions. VisionMask is trained in a self-supervised manner without relying on human-generated labels. Importantly, its training does not alter the agent model, hence preserving the agent's performance and integrity. We evaluate VisionMask on Super Mario Bros (SMB) and three Atari games. Compared to existing methods, VisionMask achieves a 14.9% higher insertion accuracy and a 30.08% higher F1-Score in reproducing original actions from the selected visual explanations. We also present examples illustrating how VisionMask can be used for counterfactual analysis.

Combinatorial mechanical metamaterials are made of anisotropic, flexible blocks, such that multiple metamaterials may be constructed using a single block type, and the system's response strongly depends on the mutual orientations of the blocks within the lattice. We study a family of possible block types for the square, honeycomb, and cubic lattices. Blocks that are centrally symmetric induce holographic order, such that mechanical compatibility (meaning that blocks do not impede each other's motion) implies bulk-boundary coupling. With them, one can design a compatible metamaterial that will deform in any desired texture only on part of its boundary. With blocks that break holographic order, we demonstrate how to design the deformation texture on the entire boundary. Correspondingly, the number of compatible holographic metamaterials scales exponentially with the boundary, while in non-holographic cases we show that it scales exponentially with the bulk., Comment: See accompanying paper: Defect Positioning in Combinatorial Metamaterials arXiv:2412.01227

We study the maximum ratio of the Euclidean norm to the comass norm of p-covectors in Euclidean n-space and improve the known upper bound found in the standard references by Whitney and Federer. We go on to prove stable systolic inequalities when the fundamental cohomology class of the manifold is a cup product of forms of lower degree., Comment: 11 pages. To appear in Differential Geometry and Its Applications

The UV continuum slope of galaxies, $\beta$, is a powerful diagnostic. Understanding the redshift evolution of $\beta$ and its dependence on key galaxy properties can shed light on the evolution of galaxy physical properties over cosmic time. In this study, we present $\beta$ measurements for 295 spectroscopically confirmed galaxies at $5.5 9.5$ to begin to redden, deviating from the trend observed at $z < 9.5$. By producing stacked spectra in bins of redshift and $\beta$, we derive trends between $\beta$ and dust attenuation, metallicity, ionization parameter, and stellar age indicators directly from spectra, finding a lack of dust attenuation to be the dominant driver of bluer $\beta$ values. We further report six galaxies with $\beta<-3.0$, which show a range of spectroscopic properties and signs of significant LyC photon leakage. Finally, we show that the redder $\beta$ values at $z > 9.5$ may require rapid build-up of dust reservoirs in the very early Universe or a significant contribution from the nebular continuum emission to the observed UV spectra, with the nebular continuum fraction depending on the gas temperatures and densities. Our modeling shows that in the absence of dust, nebular emission at $T > 15,000$ K can reproduce the range of $\beta$ that we see in our sample. Higher gas temperatures driven by hot, massive stars can boost the fraction of nebular continuum emission, potentially explaining the observed $\beta$ values as well as bright UV magnitudes seen across galaxies at $z > 10$., Comment: submitted to MNRAS, 23 pages including 1 appendix, 12 Figures (including 1 in appendix), 4 tables, comments welcome! Updated version includes median and standard deviation values in bins of redshift, MUV and beta, along with a new table (Table 2) outlining the sample splitting parameters

In this work we investigate discrete-time transport in a generic U(1)-symmetric disordered model tuned across an array of different dynamical regimes. We develop an aggregate quantity, a circular statistical moment, which is a simple function of the magnetization profile and which elegantly captures transport properties of the system. From this quantity we extract transport exponents, revealing behaviors across the phase diagram consistent with localized, diffusive, and - most interestingly for a disordered system - superdiffusive regimes. Investigation of this superdiffusive regime reveals the existence of a prethermal "swappy" regime unique to discrete-time systems in which excitations propagate coherently; even in the presence of strong disorder., Comment: Comments are welcome!

Stellar evolution is driven by the changing composition of a star from nuclear reactions. At the late stages of evolution and during explosive events, the timescale can be short and drive strong hydrodynamic flows, making simulations of astrophysical reacting flows challenging. Over the past decades, the standard approach to modeling reactions in simulation codes has been operator splitting, using implicit integrators for reactions. Here we explore some of the assumptions in this standard approach and describe some techniques for improving the efficiency and accuracy of astrophysical reacting flows., Comment: submitted to Proceedings of AstroNum 2024

Confinement prohibits isolation of color charges, e.g., quarks, in nature via a process called string breaking: the separation of two charges results in an increase in the energy of a color flux, visualized as a string, connecting those charges. Eventually, creating additional charges is energetically favored, hence breaking the string. Such a phenomenon can be probed in simpler models, including quantum spin chains, enabling enhanced understanding of string-breaking dynamics. A challenging task is to understand how string breaking occurs as time elapses, in an out-of-equilibrium setting. This work establishes the phenomenology of dynamical string breaking induced by a gradual increase of string tension over time. It, thus, goes beyond instantaneous quench processes and enables tracking the real-time evolution of strings in a more controlled setting. We focus on domain-wall confinement in a family of quantum Ising chains. Our results indicate that, for sufficiently short strings and slow evolution, string breaking can be described by the transition dynamics of a two-state quantum system akin to a Landau-Zener process. For longer strings, a more intricate spatiotemporal pattern emerges: the string breaks by forming a superposition of bubbles (domains of flipped spins of varying sizes), which involve highly excited states. We finally demonstrate that string breaking driven only by quantum fluctuations can be realized in the presence of sufficiently long-ranged interactions. This work holds immediate relevance for studying string breaking in quantum-simulation experiments., Comment: 19 pages, 15 figures

The evolution of wireless communication systems will be fundamentally impacted by an open radio access network (O-RAN), a new concept defining an intelligent architecture with enhanced flexibility, openness, and the ability to slice services more efficiently. For all its promises, and like any technological advancement, O-RAN is not without risks that need to be carefully assessed and properly addressed to accelerate its wide adoption in future mobile networks. In this paper, we present an in-depth security analysis of the O-RAN architecture, discussing the potential threats that may arise in the different O-RAN architecture layers and their impact on the Confidentiality, Integrity, and Availability (CIA) triad. We also promote the potential of zero trust, Moving Target Defense (MTD), blockchain, and large language models(LLM) technologies in fortifying O-RAN's security posture. Furthermore, we numerically demonstrate the effectiveness of MTD in empowering robust deep reinforcement learning methods for dynamic network slice admission control in the O-RAN architecture. Moreover, we examine the effect of explainable AI (XAI) based on LLMs in securing the system., Comment: 10 pages

We present the Brain Treebank, a large-scale dataset of electrophysiological neural responses, recorded from intracranial probes while 10 subjects watched one or more Hollywood movies. Subjects watched on average 2.6 Hollywood movies, for an average viewing time of 4.3 hours, and a total of 43 hours. The audio track for each movie was transcribed with manual corrections. Word onsets were manually annotated on spectrograms of the audio track for each movie. Each transcript was automatically parsed and manually corrected into the universal dependencies (UD) formalism, assigning a part of speech to every word and a dependency parse to every sentence. In total, subjects heard over 38,000 sentences (223,000 words), while they had on average 168 electrodes implanted. This is the largest dataset of intracranial recordings featuring grounded naturalistic language, one of the largest English UD treebanks in general, and one of only a few UD treebanks aligned to multimodal features. We hope that this dataset serves as a bridge between linguistic concepts, perception, and their neural representations. To that end, we present an analysis of which electrodes are sensitive to language features while also mapping out a rough time course of language processing across these electrodes. The Brain Treebank is available at https://BrainTreebank.dev/, Comment: 36 pages, 17 figures; Accepted at NeurIPS Dataset and Benchmarks 2024

Objectives: Age and gender estimation is crucial for various applications, including forensic investigations and anthropological studies. This research aims to develop a predictive system for age and gender estimation in living individuals, leveraging dental measurements such as Coronal Height (CH), Coronal Pulp Cavity Height (CPCH), and Tooth Coronal Index (TCI). Methods: Machine learning models were employed in our study, including Cat Boost Classifier (Catboost), Gradient Boosting Machine (GBM), Ada Boost Classifier (AdaBoost), Random Forest (RF), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGB), and Extra Trees Classifier (ETC), to analyze dental data from 862 living individuals (459 males and 403 females). Specifically, periapical radiographs from six teeth per individual were utilized, including premolars and molars from both maxillary and mandibular. A novel ensemble learning technique was developed, which uses multiple models each tailored to distinct dental metrics, to estimate age and gender accurately. Furthermore, an explainable AI model has been created utilizing SHAP, enabling dental experts to make judicious decisions based on comprehensible insight. Results: The RF and XGB models were particularly effective, yielding the highest F1 score for age and gender estimation. Notably, the XGB model showed a slightly better performance in age estimation, achieving an F1 score of 73.26%. A similar trend for the RF model was also observed in gender estimation, achieving a F1 score of 77.53%. Conclusions: This study marks a significant advancement in dental forensic methods, showcasing the potential of machine learning to automate age and gender estimation processes with improved accuracy.

We consider quantum networks, where entangled photon pairs are distributed using fibre optic links from a centralized source to entangling nodes. The entanglement is then stored (via an entanglement swap) in entangling nodes' quantum memories until used in, e.g., distributed quantum computing, quantum key distribution, quantum sensing, and other applications. Due to the fibre loss, some photons are lost in transmission. Noise in the transmission link and the quantum memory also reduces fidelity. Thus, entangling nodes must keep updated records of photon-pair arrivals to each destination, and their use by the applications. This coordination requires classical information exchange between each entangled node pair. However, the same fibre link may not admit both classical and quantum transmissions, as the classical channels can generate enough noise (i.e., via spontaneous Raman scattering) to make the quantum link unusable. Here, we consider coordinating entanglement distribution using a standard Internet protocol (IP) network instead, and propose a control protocol to enable such. We analyse the increase in latency from transmission over an IP network, together with the effect of photon loss, quantum memory noise and buffer size, to determine the fidelity and rate of entangled pairs. We characterize the relationship between the latency of the non-ideal IP network and the decoherence time of the quantum memories, providing a comparison of promising quantum memory technologies.

We introduce MEGATRON, a new galaxy formation model for cosmological radiation hydrodynamics simulations of high-redshift galaxies. The model accounts for the non-equilibrium chemistry and heating/cooling processes of $\geq 80$ atoms, ions, and molecules, coupled to on-the-fly radiation transfer. We apply the model in a cosmological setting to the formation of a $10^9\ {\rm M_{\odot}}$ halo at $z=6$, and run 25 realizations at pc-scale resolution, varying numerous parameters associated with our state-of-the-art star formation, stellar feedback, and chemical enrichment models. We show that the overall budget of feedback energy is the key parameter that controls star formation regulation at high redshift, with other numerical parameters (e.g. supernova clustering, star formation conditions) having a more limited impact. As a similar feedback model has been shown to produce realistic $z=0$ galaxies, our work demonstrates that calibration at $z=0$ does not guarantee strong regulation of star formation at high-redshift. Interestingly, we find that subgrid model variations that have little impact on the final $z=6$ stellar mass can lead to substantial changes on the observable properties of high-redshift galaxies. For example, different star formation models based on, e.g. density thresholds or turbulence inspired criteria, lead to fundamentally distinct nebular emission line ratios across the interstellar medium (ISM). These results highlight the ISM as an important resource for constraining models of star formation, feedback, and galaxy formation in the JWST era, where emission line measurements for $>1,000$ high-redshift galaxies are now available., Comment: 25 pages, 16 Figures, submitted to the Open Journal of Astrophysics

Among the most puzzling early discoveries of JWST are "Little Red Dots" -- compact red sources that host broad Balmer emission lines and, in many cases, exhibit a "V shaped" change in slope in the rest-optical. The physical properties of Little Red Dots currently have order-of-magnitude uncertainties, because models to explain the continuum of these sources differ immensely. Here, we leverage the complete selection of red sources in the RUBIES program, supplemented with public PRISM spectra, to study the origin of this "V shape". By fitting a broken power law with a flexible inflection point, we find that a large fraction (20/44, nearly all spatially unresolved) of extremely red H$\alpha$ emitters at $2

When evaluating stimuli reconstruction results it is tempting to assume that higher fidelity text and image generation is due to an improved understanding of the brain or more powerful signal extraction from neural recordings. However, in practice, new reconstruction methods could improve performance for at least three other reasons: learning more about the distribution of stimuli, becoming better at reconstructing text or images in general, or exploiting weaknesses in current image and/or text evaluation metrics. Here we disentangle how much of the reconstruction is due to these other factors vs. productively using the neural recordings. We introduce BrainBits, a method that uses a bottleneck to quantify the amount of signal extracted from neural recordings that is actually necessary to reproduce a method's reconstruction fidelity. We find that it takes surprisingly little information from the brain to produce reconstructions with high fidelity. In these cases, it is clear that the priors of the methods' generative models are so powerful that the outputs they produce extrapolate far beyond the neural signal they decode. Given that reconstructing stimuli can be improved independently by either improving signal extraction from the brain or by building more powerful generative models, improving the latter may fool us into thinking we are improving the former. We propose that methods should report a method-specific random baseline, a reconstruction ceiling, and a curve of performance as a function of bottleneck size, with the ultimate goal of using more of the neural recordings., Comment: 23 pages, 16 figures, Accepted at NeurIPS 2024

We demonstrate that architectures which traditionally are considered to be ill-suited for a task can be trained using inductive biases from another architecture. Networks are considered untrainable when they overfit, underfit, or converge to poor results even when tuning their hyperparameters. For example, plain fully connected networks overfit on object recognition while deep convolutional networks without residual connections underfit. The traditional answer is to change the architecture to impose some inductive bias, although what that bias is remains unknown. We introduce guidance, where a guide network guides a target network using a neural distance function. The target is optimized to perform well and to match its internal representations, layer-by-layer, to those of the guide; the guide is unchanged. If the guide is trained, this transfers over part of the architectural prior and knowledge of the guide to the target. If the guide is untrained, this transfers over only part of the architectural prior of the guide. In this manner, we can investigate what kinds of priors different architectures place on untrainable networks such as fully connected networks. We demonstrate that this method overcomes the immediate overfitting of fully connected networks on vision tasks, makes plain CNNs competitive to ResNets, closes much of the gap between plain vanilla RNNs and Transformers, and can even help Transformers learn tasks which RNNs can perform more easily. We also discover evidence that better initializations of fully connected networks likely exist to avoid overfitting. Our method provides a mathematical tool to investigate priors and architectures, and in the long term, may demystify the dark art of architecture creation, even perhaps turning architectures into a continuous optimizable parameter of the network., Comment: Under Review; 24 pages, 9 figures; Project page and code is at https://untrainable-networks.github.io/

Network verification (NWV), broadly defined as the verification of properties of distributed protocols used in network systems, cannot be efficiently solved on classical hardware via brute force. Prior work has developed a variety of methods that scale by observing a structure in the search space and then evaluating classes within the search space instead of individual instances. However, even these classification mechanisms have their limitations. In this paper, we consider a radically different approach: applying quantum computing to more efficiently solve NWV problems. We provide an overview of how to map variants of NWV problems into unstructured search problems that can be solved via quantum computing with quadratic speedup, making the approach feasible in theory to problems that are double in size (of the input). Emerging quantum systems cannot yet tackle problems of practical interest, but rapid advances in hardware and algorithm development make now a great time to start thinking about their application. With this in mind, we explore the limits of scale of the problem for which quantum computing can solve NWV problems as unstructured search.

Reliable stellar age estimates are fundamental for testing several problems in modern astrophysics, in particular since they set the time scales of Galactic dynamical and chemical evolution. In this study, we determine ages using only Gaia DR3 photometry and parallaxes, in combination with interstellar extinction maps, spectroscopic metallicities and $\alpha$ abundances from the latest data release (DR8) of the LAMOST survey. In contrast with previous age estimates, we do not use spectroscopic effective temperatures or surface gravities, thus relying on the excellent precision and accuracy of the Gaia photometry. We use a new version of the publicly available SPInS code with improved features, including the on-the-fly computation of the autocorrelation time and the automatic convergence evaluation. We determine reliable age estimates for 35,096 and 243,768 sub-giant and main-sequence turn-off stars in the LAMOST DR8 low- and medium-resolution surveys with typical uncertainties smaller than 10%. In addition, we successfully test our method on more than 4,000 stars of 14 well-studied open and globular star clusters covering a wide range of ages, confirming the reliability of our age and uncertainty estimates., Comment: Accepted for publication in A&A

Many extensions of the Standard Model propose the existence of new particles or forces, aiming to answer mysteries such as the identity of the elusive dark matter. Atomic-based detectors are at the forefront of technologies designed to search for these particles or forces through their couplings to fermions, enabling the testing of well-motivated models, such as axion-like particles, which could form dark matter. These detectors also probe new long-range interactions between the detectors and spin-polarized objects, as well as interactions mediated by light particles that break CP symmetry, introducing a coupling between the detector and an unpolarized object. However, the sensitivity of these detectors is often constrained by magnetic noise, limiting their effectiveness to a narrow region of parameter space. We propose and develop a technique, which we name the Rotating Wave comagnetometer (RoW comag), that can suppress magnetic noise at tunable frequencies while maintaining high sensitivity to target signals, significantly expanding the potential reach of these detectors. We analyze its operation for testing various extensions to the Standard Model and show how it could improve current sensitivities by several orders of magnitude. This work paves the way for a new class of tabletop experiments aimed at searching for new physics, including the exploration of well-motivated axion-like particle dark matter models at higher masses than previously attainable., Comment: 14 pages, 3 figures

The first few cycles of JWST have identified an overabundance of UV-bright galaxies and a general excess of UV luminosity density at $z\gtrsim10$ compared to expectations from most (pre-JWST) theoretical models. Moreover, some of the brightest high-redshift spectroscopically confirmed galaxies exhibit peculiar chemical abundance patterns, most notably extremely high N/O ratios. Since N/O has been empirically shown to scale strongly with He/H, as expected for hot hydrogen burning, these same bright high-redshift galaxies are likely also helium-enhanced. Under simplistic assumptions for stellar evolution, the bolometric luminosity of a star scales as $L\propto(2-\frac{5}{4}Y)^{-4}(2-Y)^{-1}$ -- hence a higher He/H leads to brighter stars. In this Letter, we evolve a series of MESA models to the zero-age main-sequence and highlight that the helium enhancements at the levels measured and inferred for high-redshift galaxies can boost the 1500 $\mathring{\rm A}$ UV luminosity by up to $\sim50\%$, while simultaneously increasing the stellar effective temperature. The combination of helium enhancements with nebular continuum emission expected for intense bursts of star formation have the potential to help reduce the tension between JWST observations and certain galaxy formation models., Comment: Note that this is an erratum, fixing an error in equation 2 where we erroneously dropped the exponent of $\mu$. All conclusions and subsequent calculations remain unchanged and corrected equations are now provided

Spontaneous particle-pair formation is a fundamental phenomenon in nature. It can, for example, appear when the potential energy between two particles increases with separation, as if they were connected by a tense string. Beyond a critical separation, new particle pairs can form, causing the string to break. String-breaking dynamics in quantum chromodynamics play a vital role in high-energy particle collisions and early universe evolution. Simulating string evolution and hadron formation is, therefore, a grand challenge in modern physics. Quantum simulators, well-suited for studying dynamics, are expected to outperform classical computing methods. However, the required experimental capabilities to simulate string-breaking dynamics have not yet been demonstrated, even for simpler models of the strong force. We experimentally probe, for the first time, the spatiotemporal dynamics of string-breaking in a (1+1)-dimensional $\mathbb{Z}_2$ lattice gauge theory using a fully programmable trapped-ion quantum simulator. We emulate external static charges and strings via site-dependent magnetic-field control enabled by a dual array of tightly focused laser beams targeting individual ions. First, we study how confinement affects isolated charges, finding that they freely spread without string tension but exhibit localized oscillations when tension is increased. Then, we observe and characterize string-breaking dynamics of a string stretched between two static charges after an abrupt increase in string tension. Charge pairs appear near the string edges and spread into the bulk, revealing a route to dynamical string-breaking distinct from the conventional Schwinger mechanism. Our work demonstrates that analog quantum simulators have achieved the necessary control to explore string-breaking dynamics, which may ultimately be relevant to nuclear and high-energy physics., Comment: 22 pages, 3 main figures, 9 extended data figures

Quantum simulation is one of the methods that have been proposed and used in practice to bypass computational challenges in the investigation of lattice gauge theories. While most of the proposals rely on truncating the infinite dimensional Hilbert spaces that these models feature, we propose a truncation-free method based on the exact analogy between the local Hilbert space of lattice QED and that of a Josephson junction. We provide several proposals, mostly semi-analog, arranged according to experimental difficulty. Our method can simulate a quasi-2D system of up to $2\times N$ plaquettes, and we present an approximate method that can simulate the fully-2D theory, but is more demanding experimentally and not immediately feasible. This sets the ground for analog quantum simulation of lattice gauge theories with superconducting circuits, in a completely Hilbert space truncation-free procedure, for continuous gauge groups., Comment: 12 pages, 6 figures. v2: added numerical estimation of errors

The training data for LLMs embeds societal values, increasing their familiarity with the language's culture. Our analysis found that 44% of the variance in the ability of GPT-4o to reflect the societal values of a country, as measured by the World Values Survey, correlates with the availability of digital resources in that language. Notably, the error rate was more than five times higher for the languages of the lowest resource compared to the languages of the highest resource. For GPT-4-turbo, this correlation rose to 72%, suggesting efforts to improve the familiarity with the non-English language beyond the web-scraped data. Our study developed one of the largest and most robust datasets in this topic area with 21 country-language pairs, each of which contain 94 survey questions verified by native speakers. Our results highlight the link between LLM performance and digital data availability in target languages. Weaker performance in low-resource languages, especially prominent in the Global South, may worsen digital divides. We discuss strategies proposed to address this, including developing multilingual LLMs from the ground up and enhancing fine-tuning on diverse linguistic datasets, as seen in African language initiatives.

In this work, we propose a Multiple Access Control (MAC) protocol for Light-based IoT (LIoT) networks, where the gateway node orchestrates and schedules batteryless nodes duty-cycles based on their location and sleep time. The LIoT concept represents a sustainable solution for massive indoor IoT applications, offering an alternative communication medium through Visible Light Communication (VLC). While most existing scheduling algorithms for intermittent batteryless IoT aim to maximize data collection and enhance dataset size, our solution is tailored for environmental sensing applications, such as temperature, humidity, and air quality monitoring, optimizing measurement distribution and minimizing blind spots to achieve comprehensive and uniform environmental sensing. We propose a Balanced Space and Time-based Time Division Multiple Access scheduling (BST-TDMA) algorithm, which addresses environmental sensing challenges by balancing spatial and temporal factors to improve the environmental sensing efficiency of batteryless LIoT nodes. Our measurement-based results show that BST-TDMA was able to efficiently schedule duty-cycles with given intervals.

The ability of large language models (LLMs) to execute complex instructions is essential for their real-world applications. However, several recent studies indicate that LLMs struggle with challenging instructions. In this paper, we propose Evolutionary Contrastive Distillation (ECD), a novel method for generating high-quality synthetic preference data designed to enhance the complex instruction-following capability of language models. ECD generates data that specifically illustrates the difference between a response that successfully follows a set of complex instructions and a response that is high-quality, but nevertheless makes some subtle mistakes. This is done by prompting LLMs to progressively evolve simple instructions to more complex instructions. When the complexity of an instruction is increased, the original successful response to the original instruction becomes a "hard negative" response for the new instruction, mostly meeting requirements of the new instruction, but barely missing one or two. By pairing a good response with such a hard negative response, and employing contrastive learning algorithms such as DPO, we improve language models' ability to follow complex instructions. Empirically, we observe that our method yields a 7B model that exceeds the complex instruction-following performance of current SOTA 7B models and is competitive even with open-source 70B models.

Nano-patterned magnetic materials have opened new venues on the investigation of strongly correlated phenomena including artificial spin-ice systems, geometric frustration, magnetic monopoles, for technologically important applications such as reconfigurable ferromagnetism. With the advent of atomically thin two-dimensional (2D) van der Waals (vdW) magnets a pertinent question is whether such compounds could make their way into this realm where interactions can be tailored so that unconventional states of matter could be assessed. Here we show that square islands of CrGeTe3 vdW ferromagnets distributed in a grid manifest antiferromagnetic correlations, essential to enable frustration resulting in an artificial spin-ice. By using a combination of SQUID-on-tip microscopy, focused ion beam lithography, and atomistic spin dynamic simulations, we show that pristine, isolated CGT flakes as small as 150*150*60 nm3 have tunable dipole-dipole interactions, which can be precisely controlled by their lateral spacing. There is a crossover between non-interacting islands and significant inter-island anticorrelation depending how they are spatially distributed allowing the creation of complex magnetic patterns not observable at the isolated flakes. Our findings suggest that the cross-talk between the nano-patterned magnets can be explored in the generation of even more complex spin configurations where exotic interactions may be manipulated in an unprecedent way., Comment: Main text: 12 pages, 5 figures. Supplementary information: 11 pages, 5 figures

There is an increasing body of work using Large Language Models (LLMs) as agents for orchestrating workflows and making decisions in domains that require planning and multi-step reasoning. As a result, it is imperative to evaluate LLMs on core skills required for planning. In this work, we present ACPBench, a benchmark for evaluating the reasoning tasks in the field of planning. The benchmark consists of 7 reasoning tasks over 13 planning domains. The collection is constructed from planning domains described in a formal language. This allows us to synthesize problems with provably correct solutions across many tasks and domains. Further, it allows us the luxury of scale without additional human effort, i.e., many additional problems can be created automatically. Our extensive evaluation of 22 LLMs and OpenAI o1 reasoning models highlights the significant gap in the reasoning capability of the LLMs. Our findings with OpenAI o1, a multi-turn reasoning model, reveal significant gains in performance on multiple-choice questions, yet surprisingly, no notable progress is made on boolean questions. The ACPBench collection is available at https://ibm.github.io/ACPBench., Comment: Added OpenAI o1 results

While "complexity science" has achieved significant successes in several interdisciplinary fields such as economics and biology, it is only a very recent observation that legal systems -- from the way legal texts are drafted and connected to the rest of the corpus, up to the level of how judges and courts reach decisions under a variety of conflicting inputs -- share several features with standard Complex Adaptive Systems. This review is meant as a gentle introduction to the use of quantitative tools and techniques of complexity science to describe, analyse, and tame the complex web of human interactions that the Law is supposed to regulate. We offer an overview of the main directions of research undertaken so far as well as an outlook for future research, and we argue that statistical physicists and complexity scientists should not ignore the opportunities offered by the cross-fertilisation between legal scholarship and complex-systems modelling., Comment: 7 pages, 4 figures, 2 tables - submitted as EPL Perspective

Dwarf galaxies hold the key to crucial frontiers of astrophysics, however, their faintness renders spectroscopy challenging. Here we present the JWST Cycle 2 survey, All the Little Things (ALT, PID 3516), which is designed to seek late-forming Pop III stars and the drivers of reionization at $z\sim6-7$. ALT has acquired the deepest NIRCam grism spectroscopy yet (7-27 hr), at JWST's most sensitive wavelengths (3-4 $\mu$m), covering the powerful lensing cluster Abell 2744. Over the same 30 arcmin$^2$, ALT's ultra-deep F070W+F090W imaging ($\sim$30 mag) enables selection of very faint sources at $z>6$. We demonstrate the success of ALT's novel ``butterfly" mosaic to solve spectral confusion and contamination, and introduce the ``Allegro" method for emission line identification. By collecting spectra for every source in the field of view, ALT has measured precise ($R\sim1600$) redshifts for 1630 sources at $z=0.2-8.5$. This includes one of the largest samples of distant dwarf galaxies: [1015, 475, 50] sources less massive than the SMC, Fornax, and Sculptor with $\log(M_{*}/M_{\odot})<$[8.5, 7.5, 6.5]. We showcase ALT's discovery space with: (i) spatially resolved spectra of lensed clumps in galaxies as faint as $M_{\rm{UV}}\sim-15$; (ii) large-scale clustering -- overdensities at $z$=[2.50, 2.58, 3.97, 4.30, 5.66, 5.77, 6.33] hosting massive galaxies with striking Balmer breaks; (iii) small-scale clustering -- a system of satellites around a Milky Way analog at $z\sim6$; (iv) spectroscopically confirmed multiple images that help constrain the lensing model underlying all science in this legacy field; (v) sensitive star-formation maps based on dust-insensitive tracers such as Pa$\alpha$; (vi) direct spectroscopic discovery of rare sources such as AGN with ionized outflows. These results provide a powerful proof of concept for how grism surveys maximize the potential of strong lensing fields., Comment: Submitted to the Open Journal of Astrophysics. Figs. 6 and 10 illustrate the quality of the spectra and imaging, while Fig. 12 summarizes the yield of the survey. Comments warmly welcomed and greatly appreciated. The ALT DR1 catalog is available at https://zenodo.org/records/13871850