Search

Kernel-based random graphs (KBRGs) are a broad class of random graph models that account for inhomogeneity among vertices. We consider KBRGs on a discrete $d-$dimensional torus $\mathbf{V}_N$ of size $N^d$. Conditionally on an i.i.d.~sequence of weights $(W_i)_{i\in \mathbf{V}_N}$ with tail exponent $\tau-1>0$, we connect any two points $i$ and $j$ on the torus with probability $$p_{ij}= \frac{\kappa_{\sigma}(W_i,W_j)}{\|i-j\|^{\alpha}} \wedge 1$$ for some parameter $\alpha>0$ and $\kappa_{\sigma}(u,v)= (u\vee v)(u \wedge v)^{\sigma}$ for some $\sigma\in(0,\tau-1)$. We focus on the adjacency operator of this random graph and study its empirical spectral distribution. For $\alpha2$, we show that a non-trivial limiting distribution exists as $N\to\infty$. The corresponding measure is absolutely continuous with respect to the Lebesgue measure and we analyse its moments. The measure is given by an operator-valued semicircle law, whose Stieltjes transform is characterised by a fixed point equation in an appropriate Banach space. In the case $\sigma=1$, corresponding to the so-called scale-free percolation random graph, we can explicitly describe the limiting measure and study its tail., Comment: 46 pages, 6 figures

Heterozygous mutations in KMT2B are associated with an early-onset, progressive, and often complex dystonia (DYT28). Key characteristics of typical disease include focal motor features at disease presentation, evolving through a caudocranial pattern into generalized dystonia, with prominent oromandibular, laryngeal, and cervical involvement. Although KMT2B-related disease is emerging as one of the most common causes of early-onset genetic dystonia, much remains to be understood about the full spectrum of the disease. We describe a cohort of 53 patients with KMT2B mutations, with detailed delineation of their clinical phenotype and molecular genetic features. We report new disease presentations, including atypical patterns of dystonia evolution and a subgroup of patients with a non-dystonic neurodevelopmental phenotype. In addition to the previously reported systemic features, our study has identified co-morbidities, including the risk of status dystonicus, intrauterine growth retardation, and endocrinopathies. Analysis of this study cohort (n = 53) in tandem with published cases (n = 80) revealed that patients with chromosomal deletions and protein-truncating variants had a significantly higher burden of systemic disease (with earlier onset of dystonia) than those with missense variants. Eighteen individuals had detailed longitudinal data available after insertion of deep brain stimulation for medically refractory dystonia. Median age at deep brain stimulation was 11.5 years (range: 4.5 to 37.0 years). Follow-up after deep brain stimulation ranged from 0.25 to 22 years. Significant improvement of motor function and disability (as assessed by the Burke-Fahn-Marsden Dystonia Rating Scales, BFMDRS-M and BFMDRS-D) was evident at 6 months, 1 year, and last follow-up (motor, P = 0.001, P = 0.004, and P = 0.012; disability, P = 0.009, P = 0.002, and P = 0.012)., Comment: PMID: 33150406; PMCID: PMC7719027

Self-Explainable Graph Neural Networks (SE-GNNs) are popular explainable-by-design GNNs, but the properties and the limitations of their explanations are not well understood. Our first contribution fills this gap by formalizing the explanations extracted by SE-GNNs, referred to as Trivial Explanations (TEs), and comparing them to established notions of explanations, namely Prime Implicant (PI) and faithful explanations. Our analysis reveals that TEs match PI explanations for a restricted but significant family of tasks. In general, however, they can be less informative than PI explanations and are surprisingly misaligned with widely accepted notions of faithfulness. Although faithful and PI explanations are informative, they are intractable to find and we show that they can be prohibitively large. Motivated by this, we propose Dual-Channel GNNs that integrate a white-box rule extractor and a standard SE-GNN, adaptively combining both channels when the task benefits. Our experiments show that even a simple instantiation of Dual-Channel GNNs can recover succinct rules and perform on par or better than widely used SE-GNNs. Our code can be found in the supplementary material.

The rapid adoption of pervasive and mobile computing has led to an unprecedented rate of data production and consumption by mobile applications at the network edge. These applications often require interactions such as data exchange, behavior coordination, and collaboration, which are typically mediated by cloud servers. While cloud computing has been effective for distributed systems, challenges like latency, cost, and intermittent connectivity persist. With the advent of 5G technology, features like location-awareness and device-to-device (D2D) communication enable a more distributed and adaptive architecture. This paper introduces Self-Organizing Interaction Spaces (SOIS), a novel framework for engineering pervasive applications. SOIS leverages the dynamic and heterogeneous nature of mobile nodes, allowing them to form adaptive organizational structures based on their individual and social contexts. The framework provides two key abstractions for modeling and programming pervasive applications using an organizational mindset and mechanisms for adapting dynamic organizational structures. Case examples and performance evaluations of a simulated mobile crowd-sensing application demonstrate the feasibility and benefits of SOIS. Results highlight its potential to enhance efficiency and reduce reliance on traditional cloud models, paving the way for innovative solutions in mobile and distributed environments., Comment: 9 pages, 3 listings

This report investigates the application of deep reinforcement learning (DRL) algorithms for dynamic resource allocation in wireless communication systems. An environment that includes a base station, multiple antennas, and user equipment is created. Using the RLlib library, various DRL algorithms such as Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) are then applied. These algorithms are compared based on their ability to optimize resource allocation, focusing on the impact of different learning rates and scheduling policies. The findings demonstrate that the choice of algorithm and learning rate significantly influences system performance, with DRL providing more efficient resource allocation compared to traditional methods., Comment: 6 pages, 8 figures

Distributed File Systems (DFS) are essential for managing vast datasets across multiple servers, offering benefits in scalability, fault tolerance, and data accessibility. This paper presents a comprehensive evaluation of three prominent DFSs - Google File System (GFS), Hadoop Distributed File System (HDFS), and MinIO - focusing on their fault tolerance mechanisms and scalability under varying data loads and client demands. Through detailed analysis, how these systems handle data redundancy, server failures, and client access protocols, ensuring reliability in dynamic, large-scale environments is assessed. In addition, the impact of system design on performance, particularly in distributed cloud and computing architectures is assessed. By comparing the strengths and limitations of each DFS, the paper provides practical insights for selecting the most appropriate system for different enterprise needs, from high availability storage to big data analytics., Comment: 9 pages, 3 figures, 3 tables

This paper represents "Cloudlab", a comprehensive, cloud - native laboratory designed to support network security research and training. Built on Google Cloud and adhering to GitOps methodologies, Cloudlab facilitates the the creation, testing, and deployment of secure, containerized workloads using Kubernetes and serverless architectures. The lab integrates tools like Palo Alto Networks firewalls, Bridgecrew for "Security as Code," and automated GitHub workflows to establish a robust Continuous Integration/Continuous Machine Learning pipeline. By providing an adaptive and scalable environment, Cloudlab supports advanced security concepts such as role-based access control, Policy as Code, and container security. This initiative enables data scientists and engineers to explore cutting-edge practices in a dynamic cloud-native ecosystem, fostering innovation and improving operational resilience in modern IT infrastructures., Comment: 7 pages, 5 figures

Named for orbital kinship with Pluto, the Plutinos are a prominent group of Kuiper Belt objects whose orbital periods are in libration about the 3/2 ratio with Neptune's. We investigate the long term orbital dynamics of known Plutinos, with attention to the additional libration (or lack thereof) of their argument of perihelion, $g$, a well-known characteristic of Pluto's orbit. We show that the $g$ librators amongst the Plutinos cluster around an arc in the eccentricity--inclination parameter plane. This previously unreported dynamical structure is owed to a family of periodic orbits of the third kind in the restricted problem of three bodies, identified by Poincar\'e at the end of the 19th century. Approximately sixteen percent of the currently known Plutinos exhibit $g$ librations, a far greater fraction than the ratios of the associated libration frequencies. These results may offer new constraints for theoretical models of the dynamical history of the Plutinos and of the orbital migration history of the giant planets., Comment: Accepted for publication in ApJ. 9 pages

We show that the problem of the design of the lattices of elastoplastic current conducting springs with optimal multi-functional properties leads to an analytically tractable problem. Specifically, focusing on a lattice with a small number of springs, we use the technique of inequalities to reduce the number variables and to compute the minimal cost of lattice fabrication explicitly., Comment: 20 pages, 10 figures

In recent years, it has been shown empirically that standard disentangled latent variable models do not support robust compositional learning in the visual domain. Indeed, in spite of being designed with the goal of factorising datasets into their constituent factors of variations, disentangled models show extremely limited compositional generalisation capabilities. On the other hand, object-centric architectures have shown promising compositional skills, albeit these have 1) not been extensively tested and 2) experiments have been limited to scene composition -- where models must generalise to novel combinations of objects in a visual scene instead of novel combinations of object properties. In this work, we show that these compositional generalisation skills extend to this later setting. Furthermore, we present evidence pointing to the source of these skills and how they can be improved through careful training. Finally, we point to one important limitation that still exists which suggests new directions of research., Comment: As it appeared in the Compositional Learning Workshop, NeurIPS 2024; 14 pages (5 main text, 7 appendices, 2 references); 9 figures

In this work, we have implemented an accurate machine-learning approach for predicting various key analog and RF parameters of Negative Capacitance Field-Effect Transistors (NCFETs). Visual TCAD simulator and the Python high-level language were employed for the entire simulation process. However, the computational cost was found to be excessively high. The machine learning approach represents a novel method for predicting the effects of different sources on NCFETs while also reducing computational costs. The algorithm of an artificial neural network can effectively predict multi-input to single-output relationships and enhance existing techniques. The analog parameters of Double Metal Double Gate Negative Capacitance FETs (D2GNCFETs) are demonstrated across various temperatures ($T$), oxide thicknesses ($T_{ox}$), substrate thicknesses ($T_{sub}$), and ferroelectric thicknesses ($T_{Fe}$). Notably, at $T=300K$, the switching ratio is higher and the leakage current is $84$ times lower compared to $T=500K$. Similarly, at ferroelectric thicknesses $T_{Fe}=4nm$, the switching ratio improves by $5.4$ times compared to $T_{Fe}=8nm$. Furthermore, at substrate thicknesses $T_{sub}=3nm$, switching ratio increases by $81\%$ from $T_{sub}=7nm$. For oxide thicknesses at $T_{ox}=0.8nm$, the ratio increases by $41\%$ compared to $T_{ox}=0.4nm$. The analysis reveals that $T_{Fe}=4nm$, $T=300K$, $T_{ox}=0.8nm$, and $T_{sub}=3nm$ represent the optimal settings for D2GNCFETs, resulting in significantly improved performance. These findings can inform various applications in nanoelectronic devices and integrated circuit (IC) design., Comment: 7 pages, 8 figures

High-precision astrometry offers a promising approach to detect low-frequency gravitational waves, complementing pulsar timing array (PTA) observations. We explore the response of astrometric measurements to a stochastic gravitational wave background (SGWB) in synergy with PTA data. Analytical, covariant expressions for this response are derived, accounting for the presence of a possible dipolar anisotropy in the SGWB. We identify the optimal estimator for extracting SGWB information from astrometric observations and examine how sensitivity to SGWB properties varies with the sky positions of stars and pulsars. Using representative examples of current PTA capabilities and near-future astrometric sensitivity, we demonstrate that cross-correlating astrometric and PTA data can improve constraints on SGWB properties, compared to PTA data alone. The improvement is quantified through Fisher forecasts for the SGWB amplitude, spectral tilt, and dipolar anisotropy amplitude. In the future, such joint constraints could play a crucial role in identifying the origin of SGWB signals detected by PTAs., Comment: 33 pages, 10 figures

We propose a minimal Type-I Dirac seesaw which accommodates a thermal scalar dark matter (DM) candidate protected by a CP symmetry, without introducing any additional field beyond the ones taking part in the seesaw. A $Z_4$ symmetry is introduced to realise the tree level Dirac seesaw while the Majorana mass terms are prevented by an unbroken global lepton number symmetry. While the spontaneous $Z_4$ breaking together with electroweak symmetry breaking lead to the generation of light Dirac neutrino mass, it also results in the formation of domain walls. These cosmologically catastrophic walls can be made to annihilate away by introducing bias terms while also generating stochastic gravitational waves (GW) within reach of near future experiments like LISA, BBO, $\mu$-ARES etc. The scalar DM parameter space can be probed at direct and indirect search experiments. Light Dirac neutrinos also enhance the relativistic degrees of freedom $N_{\rm eff}$ within reach of future cosmic microwave background (CMB) experiments. The model can also explain the observed baryon asymmetry via Dirac leptogenesis., Comment: 36 pages, 13 captioned figures

Fused Deposition Modeling (FDM) is a widely used additive manufacturing (AM) technique valued for its flexibility and cost-efficiency, with applications in a variety of industries including healthcare and aerospace. Recent developments have made affordable FDM machines accessible and encouraged adoption among diverse users. However, the design, planning, and production process in FDM require specialized interdisciplinary knowledge. Managing the complex parameters and resolving print defects in FDM remain challenging. These technical complexities form the most critical barrier preventing individuals without technical backgrounds and even professional engineers without training in other domains from participating in AM design and manufacturing. Large Language Models (LLMs), with their advanced capabilities in text and code processing, offer the potential for addressing these challenges in FDM. However, existing research on LLM applications in this field is limited, typically focusing on specific use cases without providing comprehensive evaluations across multiple models and tasks. To this end, we introduce FDM-Bench, a benchmark dataset designed to evaluate LLMs on FDM-specific tasks. FDM-Bench enables a thorough assessment by including user queries across various experience levels and G-code samples that represent a range of anomalies. We evaluate two closed-source models (GPT-4o and Claude 3.5 Sonnet) and two open-source models (Llama-3.1-70B and Llama-3.1-405B) on FDM-Bench. A panel of FDM experts assess the models' responses to user queries in detail. Results indicate that closed-source models generally outperform open-source models in G-code anomaly detection, whereas Llama-3.1-405B demonstrates a slight advantage over other models in responding to user queries. These findings underscore FDM-Bench's potential as a foundational tool for advancing research on LLM capabilities in FDM.

We describe the software package $\texttt{ESpRESSO}$ - [E]xtragalactic [Sp]ectroscopic [R]oman [E]mulator and [S]imulator of [S]ynthetic [O]bjects, created to emulate the slitless spectroscopic observing modes of the Nancy Grace Roman Space Telescope (Roman) Wide Field Instrument (WFI). We combine archival Hubble Space Telescope (HST) imaging data of comparable spatial resolution with model spectral energy distributions to create a data-cube of flux density as a function of position and wavelength. This data-cube is used for simulating a nine detector grism observation, producing a crowded background scene which model field angle dependent optical distortions expected for the grism. We also demonstrate the ability to inject custom sources using the described tools and pipelines. In addition, we show that spectral features such as emission line pairs are unlikely to be mistaken as off order contaminating features and vice versa. Our result is a simulation suite of half of the eighteen detector array, with a realistic background scene and injected Ly$\alpha$ emitter (LAE) galaxies, realized at 25 position angles (PAs), 12 with analogous positive and negative dithers, Using an exposure time of 10ks per PA, the full PA set can be used as a mock deep Roman grism survey with high (synthetic) LAE completeness for developing future spectral data analysis tools., Comment: 15 pages, 10 figures, 1 supplementary figure, 1 table. Submitted to MNRAS. Comments are welcome!

Symmetry breaking phase transitions from less to more ordered phases will typically produce topological defects in the ordered phase. Kibble-Zurek theory predicts that for any second-order phase transition, such as the early universe, the density of defects that form should be determined by the scaling law for the system coherence time and the phase transition quench time. We have performed measurements of fourth sound dissipation due to vortex mutual friction in thin channels of superfluid $^3$He where one spatial dimension is smaller than a characteristic length scale predicted by the Kibble-Zurek theory. Our measurements suggest that remnant vortices form after the normal to superfluid second-order phase transition, and that the density of vortices is correlated with the size of the channel, but crucially, is independent of quench time. We propose a modified picture of defect formation, where closely spaced walls prevent the ends of vortex lines from reconnecting into loops. This leads to a mean vortex separation set by the wall spacing, which can result in much higher defect densities than in bulk systems., Comment: 4 pages, 3 figures

Recently, the James Webb Space Telescope (JWST) has revealed a new class of high redshift (high-$z$, $z>4$) compact galaxies which are red in the rest-frame optical and blue in the rest-frame UV as V-shaped spectral energy distributions (SEDs), referred to as "Little Red Dots" (LRDs). It is very likely that LRDs host obscured broad-line active galactic nuclei (AGNs). In the meanwhile, Green pea galaxies (GPs), which are compact dwarf galaxies at low redshift, share various similar properties with high redshift star-forming galaxies. Here we aim to find the connection between the LRDs and GPs hosting broad-line AGNs (BLGPs). With a sample of 19 BLGPs obtained from our previous work, we further identify 7 GPs with V-shaped rest-frame UV-to-optical SEDs that are likely local analogs to LRDs. These V-shaped BLGPs exhibit faint UV absolute magnitudes and sub-Eddington rates similar to those of LRDs. Three of them occupy a similar region as LRDs in the BPT diagram, suggesting they have similar ionization conditions and gas-phase metallicities to LRDs. These similarities suggest that V-shaped BLGPs can be taken as local analogs of high-redshift LRDs. In addition, most (16/19) BLGPs, including 6 V-shaped BLGPs, host over-massive black holes above the local $M_{\rm BH}$-$M_{*}$ relation, making it the first sample of galaxies hosting over-massive black holes at $z<0.4$. These findings will help us learn more about the formation and co-evolution of early galaxies and black holes., Comment: 11 pages, 4 figures, 2 tables, submitted to the Astrophysical Journal Letters. Comments welcome!

This paper introduces a task-specific, model-agnostic framework for evaluating dataset similarity, providing a means to assess and compare dataset realism and quality. Such a framework is crucial for augmenting real-world data, improving benchmarking, and making informed retraining decisions when adapting to new deployment settings, such as different sites or frequency bands. The proposed framework is employed to design metrics based on UMAP topology-preserving dimensionality reduction, leveraging Wasserstein and Euclidean distances on latent space KNN clusters. The designed metrics show correlations above 0.85 between dataset distances and model performances on a channel state information compression unsupervised machine learning task leveraging autoencoder architectures. The results show that the designed metrics outperform traditional methods.

The modestly eccentric and non-coplanar orbits of the giant planets pose a challenge to solar system formation theories which generally indicate that the giant planets emerged from the protoplanetary disk in nearly perfectly circular and coplanar orbits. We demonstrate that a single encounter with a 2-50 Jupiter-mass object, passing through the solar system at a perihelion distance less than 20 AU and a hyperbolic excess velocity less than 6 km/s, can excite the giant planets' eccentricities and mutual inclinations to values comparable to those observed. We estimate that there is about a 1-in-100 chance that such a flyby produces a dynamical architecture similar to that of the solar system. We describe a metric to evaluate how closely a simulated system matches the eccentricity and inclination secular modes of the solar system. The scenario of a close encounter with a substellar object offers a plausible explanation for the origin of the moderate eccentricities and inclinations and the secular architecture of the planets., Comment: 11 pages, 6 figures

Compute-in-memory (CiM)-based binary neural network (CiM-BNN) accelerators marry the benefits of CiM and ultra-low precision quantization, making them highly suitable for edge computing. However, CiM-enabled crossbar (Xbar) arrays are plagued with hardware non-idealities like parasitic resistances and device non-linearities that impair inference accuracy, especially in scaled technologies. In this work, we first analyze the impact of Xbar non-idealities on the inference accuracy of various CiM-BNNs, establishing that the unique properties of CiM-BNNs make them more prone to hardware non-idealities compared to higher precision deep neural networks (DNNs). To address this issue, we propose BinSparX, a training-free technique that mitigates non-idealities in CiM-BNNs. BinSparX utilizes the distinct attributes of BNNs to reduce the average current generated during the CiM operations in Xbar arrays. This is achieved by statically and dynamically sparsifying the BNN weights and activations, respectively (which, in the context of BNNs, is defined as reducing the number of +1 weights and activations). This minimizes the IR drops across the parasitic resistances, drastically mitigating their impact on inference accuracy. To evaluate our technique, we conduct experiments on ResNet-18 and VGG-small CiM-BNNs designed at the 7nm technology node using 8T-SRAM and 1T-1ReRAM. Our results show that BinSparX is highly effective in alleviating the impact of non-idealities, recouping the inference accuracy to near-ideal (software) levels in some cases and providing accuracy boost of up to 77.25%. These benefits are accompanied by energy reduction, albeit at the cost of mild latency/area increase.

In this article, we study the following quasilinear Schr\"{o}dinger equation involving Hardy potential and Choquard type exponential nonlinearity with a parameter $\alpha$ \begin{equation*} \left\{ \begin{array}{l} - \Delta_N w - \Delta_N(|w|^{2\alpha}) |w|^{2\alpha - 2} w - \lambda \frac{|w|^{2\alpha N-2}w}{\left( |x| \log\left(\frac{R}{|x|} \right) \right)^N} = \left(\int_{\Omega} \frac{H(y,w(y))}{|x-y|^{\mu}}dy\right) h(x,w(x))\; \mbox{in }\; \Omega, w > 0 \mbox{ in } \Omega \setminus \{ 0\}, \quad \quad w = 0 \mbox{ on } \partial \Omega, \end{array} \right. \end{equation*} where $N\geq 2$, $\alpha>\frac12$, $0\leq \lambda< \left(\frac{N-1}{N}\right)^N$, $0 < \mu < N$, $h : \mathbb R^N \times \mathbb R \rightarrow \mathbb R$ is a continuous function with critical exponential growth in the sense of the Trudinger-Moser inequality and $H(x,t)= \int_{0}^{t} h(x,s) ds$ is the primitive of $h$. With the help of Mountain Pass Theorem and critical level which is obtained by the sequence of Moser functions, we establish the existence of a positive solution for a small range of $\lambda$. Moreover, we also investigate the existence of a positive solution for a non-homogeneous problem for every $0\leq \lambda <\left(\frac{N-1}{N}\right)^N.$ To the best of our knowledge, the results obtained here are new even in case of $N$-Laplace equation with Hardy potential.

To tackle the global challenge of online hate speech, a large body of research has developed detection models to flag hate speech in the sea of online content. Yet, due to systematic biases in evaluation datasets, detection performance in real-world settings remains unclear, let alone across geographies. To address this issue, we introduce HateDay, the first global hate speech dataset representative of social media settings, randomly sampled from all tweets posted on September 21, 2022 for eight languages and four English-speaking countries. Using HateDay, we show how the prevalence and composition of hate speech varies across languages and countries. We also find that evaluation on academic hate speech datasets overestimates real-world detection performance, which we find is very low, especially for non-European languages. We identify several factors explaining poor performance, including models' inability to distinguish between hate and offensive speech, and the misalignment between academic target focus and real-world target prevalence. We finally argue that such low performance renders hate speech moderation with public detection models unfeasible, even in a human-in-the-loop setting which we find is prohibitively costly. Overall, we emphasize the need to evaluate future detection models from academia and platforms in real-world settings to address this global challenge.

The epoch of reionization is a landmark in structure formation and galaxy evolution. How it happened is still not clear, especially regarding which population of objects was responsible for contributing the bulk of ionizing photons toward this process. Doubly-peaked Lyman-Alpha profiles in this epoch are of particular interest since they hold information about the escape of ionizing radiation and the environment surrounding the source. We wish to understand the escape mechanisms of ionizing radiation in Lyman-Alpha emitters during this time and the origin of a doubly-peaked Lyman-alpha profile as well as estimating the size of a potential ionized bubble. Using radiative transfer models, we fit the line profile of a bright Lyman-Alpha emitter at $z\sim 6.9$ using various gas geometries. The line modeling reveals significant radiation escape from this system. While the studied source reveals significant escape ($f_{esc}$(LyA) $\sim0.8$ as predicted by the best fitting radiative transfer model) and appears to inhabit an ionized bubble of radius $R_{b}\approx 0.8^{+0.5}_{-0.3}\,pMpc\left(\frac{t_{\rm age}}{10^{8}}\right)^{\frac{1}{3}}$.Radiative transfer modeling predicts the line to be completely redwards of the systemic redshift. We suggest the line morphology is produced by inflows, multiple components emitting Ly$\alpha$, or by an absorbing component in the red wing. We propose that CDFS-1's profile holds two red peaks produced by winds within the system. Its high $f_{esc}$(Lya) and the low-velocity offset from the systemic redshift suggest that the source is an active ionizing agent. Future observations will reveal whether a peak is present bluewards of the systemic redshift or if multiple components produce the profile., Comment: 10 pages, 7 figurss

We have used the Karl G. Jansky Very Large Array (VLA) to map H{\sc i} 21\,cm emission from the Green Pea galaxy GP~J1148+2546 at $z\approx0.0451$, only the second measurement of the H{\sc i} spatial distribution of a Green Pea. The VLA H{\sc i} 21\,cm image, the DECaLS optical image, and Sloan Digital Sky Survey spectroscopy show that GP~J1148+2546 has two neighbours, the nearer of which is only $\approx 17.5$~kpc away, and that the H{\sc i} 21\,cm emission extends in an inverted ``C'' shape around the Green Pea and its companions, with the highest H{\sc i} column density between the two neighbouring galaxies. The starburst in GP~J1148+2546 is likely to have been triggered by the ongoing merger with its neighbours, although the velocity field and velocity dispersion images do not show clear merger signatures at the Green Pea location. The H{\sc i} mass of the Green Pea and its immediate surroundings is $(3.58 \pm 0.37) \times 10^9 \, M_\odot$, a factor of $\approx 7.4$ lower than the total H{\sc i} mass of the system of three interacting galaxies, while the H{\sc i} depletion timescale of GP~J1148+2546 is $\approx 0.69$~Gyr, much shorter than that of typical galaxies at $z \approx 0$. We detect damped Ly$\alpha$ absorption and Ly$\alpha$ emission from the Green Pea in a Hubble Space Telescope Cosmic Origins Spectrograph spectrum, obtaining a high H{\sc i} column density, $\approx 2.0 \times 10^{21}$~cm$^{-2}$, and a low Ly$\alpha$ escape fraction, $\approx 0.8$\%, consistent with the relatively low value ($\approx 5.4$) of the ratio O32~$\equiv$~[O{\sc iii}]$\lambda 5007 + \lambda 4959$/[O{\sc ii}]$\lambda$3727,3729., Comment: 13 pages, 8 figures. Accepted for publication in The Astrophysical Journal

We present images of the Vega planetary debris disk obtained at 15.5, 23, and 25.5 microns with the Mid-Infrared Instrument (MIRI) on JWST. The debris system is remarkably symmetric and smooth, and centered accurately on the star. There is a broad Kuiper-belt-analog ring at 80 to 170 au that coincides with the planetesimal belt detected with ALMA at 1.34 mm. The interior of the broad belt is filled with warm debris that shines most efficiently at mid-infrared along with a shallow flux dip/gap at 60 au from the star. These qualitative characteristics argue against any Saturn-mass planets orbiting the star outside of about 10 au assuming the unseen planet would be embedded in the very broad planetesimal disk from a few to hundred au. We find that the distribution of dust detected interior to the broad outer belt is consistent with grains being dragged inward by the Poynting-Robertson effect. Tighter constraints can be derived for planets in specific locations, for example any planet shepherding the inner edge of the outer belt is likely to be less than 6 Earth masses. The disk surface brightness profile along with the available infrared photometry suggest a disk inner edge near 3-5 au, disconnected from the sub-au region that gives rise to the hot near-infrared excess. The gap between the hot, sub-au zone and the inner edge of the warm debris might be shepherded by a modest mass, Neptune-size planet., Comment: accepted for publication in ApJ

We consider a simple lattice spring model in which every spring is elastoplastic and is capable to conduct current. The elasticity bounds of spring $i$ are taken as $[-c_i,c_i]$ and the resistance of spring $i$ is taken as $1/c_i$, which allows us to compute the resistance of the system. The model is further subjected to a gradual stretching and, due to plasticity, the response force increases until a certain terminal value. We demonstrate that the recently developed sweeping process theory can be used to optimize the interplay between the terminal response force and the resistance on a physical domain of parameters $c_i.$ The proposed methodology can be used by practitioners for the design of multi-functional materials as an alternative to topological optimization., Comment: 21 pages, 11 figures

We propose a novel approach that uses large language models (LLMs) to generate persona-driven conversations between Players and Non-Player Characters (NPC) in games. Showcasing the application of our methodology, we introduce the Minecraft Persona-driven Dialogue dataset (MCPDial). Starting with a small seed of expert-written conversations, we employ our method to generate hundreds of additional conversations. Each conversation in the dataset includes rich character descriptions of the player and NPC. The conversations are long, allowing for in-depth and extensive interactions between the player and NPC. MCPDial extends beyond basic conversations by incorporating canonical function calls (e.g. "Call find a resource on iron ore") between the utterances. Finally, we conduct a qualitative analysis of the dataset to assess its quality and characteristics.

We use a suite of the most recent cosmological observations to test models of dynamical dark energy motivated by quantum gravity. Specifically, we focus on hilltop quintessence scenarios, able to satisfy theoretical constraints from quantum gravity. We discuss their realisation based on axions, their supersymmetric partners, and Higgs-like string constructions. We also examine a specific parameterisation for dynamical dark energy suitable for hilltop quintessence. We then perform an analysis based on Markov Chain Monte-Carlo to assess their predictions against CMB, galaxy surveys, and supernova data. We show to what extent current data can distinguish amongst different hilltop set-ups, providing model parameter constraints that are complementary to and synergetic with theoretical bounds from quantum gravity conjectures, as well as model comparisons across the main dark energy candidates in the literature. However, all these constraints are sensitive to priors based on theoretical assumptions about viable regions of parameter space. Consequently, we discuss theoretical challenges in refining these priors, with the aim of maximizing the informative power of current and forthcoming cosmological datasets for testing dark energy scenarios in quantum gravity., Comment: 35 pages + appendices, 24 figures, 9 tables

Wearable sensors have become ubiquitous thanks to a variety of health tracking features. The resulting continuous and longitudinal measurements from everyday life generate large volumes of data; however, making sense of these observations for scientific and actionable insights is non-trivial. Inspired by the empirical success of generative modeling, where large neural networks learn powerful representations from vast amounts of text, image, video, or audio data, we investigate the scaling properties of sensor foundation models across compute, data, and model size. Using a dataset of up to 40 million hours of in-situ heart rate, heart rate variability, electrodermal activity, accelerometer, skin temperature, and altimeter per-minute data from over 165,000 people, we create LSM, a multimodal foundation model built on the largest wearable-signals dataset with the most extensive range of sensor modalities to date. Our results establish the scaling laws of LSM for tasks such as imputation, interpolation and extrapolation, both across time and sensor modalities. Moreover, we highlight how LSM enables sample-efficient downstream learning for tasks like exercise and activity recognition.

Lunar ejecta, produced by meteoroidal impacts, have been proposed for the origin of the near-Earth asteroid (469219) Kamo'oalewa, supported by its unusually Earth-like orbit and L-type reflectance spectrum (Sharkey et al., 2021). In a recent study (Castro-Cisneros et al. 2023), we found with N-body numerical simulations that the orbit of Kamo'oalewa is dynamically compatible with rare pathways of lunar ejecta captured into Earth's co-orbital region, persistently transitioning between horseshoe and quasi-satellite (HS-QS) states. Subsequently, Jiao et al. (2024) found with hydrodynamic and N-body simulations that the geologically young lunar crater Giordano Bruno generated up to 300 Kamo'oalewa-sized escaping fragments, and up to three of those could have become Earth co-orbitals. However, these results are based upon specific initial conditions of the major planets in the Solar System, close to the current epoch. In particular, over megayear time spans, Earth's eccentricity undergoes excursions up to five times its current value, potentially affecting the chaotic orbital evolution of lunar ejecta and their capture into Earth's co-orbital regions. In the present work, we carry out additional numerical simulations to compute the statistics of co-orbital outcomes across different launch epochs, representative of the full range of Earth's eccentricity values. Our main results are as follows: Kamo'oalewa-like co-orbital outcomes of lunar ejecta vary only slightly across the range of Earth's orbital eccentricity, suggesting no privileged ejecta launching epoch for such objects; the probability of co-orbital outcomes decreases rapidly with increasing launch speed, but long-lived HS-QS states are favored at higher launch speeds.

Ranking a set of items based on their relevance to a given query is a core problem in search and recommendation. Transformer-based ranking models are the state-of-the-art approaches for such tasks, but they score each query-item independently, ignoring the joint context of other relevant items. This leads to sub-optimal ranking accuracy and high computational costs. In response, we propose Cross-encoders with Joint Efficient Modeling (CROSS-JEM), a novel ranking approach that enables transformer-based models to jointly score multiple items for a query, maximizing parameter utilization. CROSS-JEM leverages (a) redundancies and token overlaps to jointly score multiple items, that are typically short-text phrases arising in search and recommendations, and (b) a novel training objective that models ranking probabilities. CROSS-JEM achieves state-of-the-art accuracy and over 4x lower ranking latency over standard cross-encoders. Our contributions are threefold: (i) we highlight the gap between the ranking application's need for scoring thousands of items per query and the limited capabilities of current cross-encoders; (ii) we introduce CROSS-JEM for joint efficient scoring of multiple items per query; and (iii) we demonstrate state-of-the-art accuracy on standard public datasets and a proprietary dataset. CROSS-JEM opens up new directions for designing tailored early-attention-based ranking models that incorporate strict production constraints such as item multiplicity and latency.

Classification tasks present challenges due to class imbalances and evolving data distributions. Addressing these issues requires a robust method to handle imbalances while effectively detecting out-of-distribution (OOD) samples not encountered during training. This study introduces a novel OOD detection algorithm designed for tabular datasets, titled Deep Neural Network-based Gaussian Descriptor for Imbalanced Tabular Data (DNN-GDITD). The DNN-GDITD algorithm can be placed on top of any DNN to facilitate better classification of imbalanced data and OOD detection using spherical decision boundaries. Using a combination of Push, Score-based, and focal losses, DNN-GDITD assigns confidence scores to test data points, categorizing them as known classes or as an OOD sample. Extensive experimentation on tabular datasets demonstrates the effectiveness of DNN-GDITD compared to three OOD algorithms. Evaluation encompasses imbalanced and balanced scenarios on diverse tabular datasets, including a synthetic financial dispute dataset and publicly available tabular datasets like Gas Sensor, Drive Diagnosis, and MNIST, showcasing DNN-GDITD's versatility., Comment: 17 pages