Search

This paper proposes the Mobile Cell (MC) concept for on-demand 5G private networks. The MC is designed to extend, restore, and reinforce 5G wireless coverage and network capacity on-demand, especially in areas with temporary communications needs or where it is costly or not possible to deploy a permanent fixed infrastructure. The design of the MC as well as the development, integration, and deployment in 5G private networks are discussed. The Mobile Cell concept can be applied in multiple real-world environments, including seaports and application scenarios. Similarly to critical hubs in the global supply chain, seaports require reliable, high-performance wireless communications to increase efficiency and manage dynamic operations in real-time. Current communications solutions in seaports typically rely on Wi-Fi and wired-based technologies. Wired-based technologies lack the necessary flexibility for dynamic environments. Wi-Fi is susceptible to interference from other systems operating in the same frequency bands. An MC operating in a licensed, interference-free spectrum is a promising solution to overcome these limitations and provide improved Quality of Service when using the 5G technology., Comment: Manuscript accepted at the NEXUS International Conference - DGTMP 2024

We present high pressure electrical transport, magnetization, and single crystal X-ray diffraction data on SrCo2P2 single crystals. X-ray diffraction data show that there is a transition to a collapsed tetragonal structure for p ~> 10 GPa and measurements of resistance show that above ~ 10 GPa, a clear transition-like feature can be observed at temperatures up to 260 K. Further magnetization, magnetoresistance and Hall measurements made under pressure all indicate that this transition is to a ferromagnetic ground state. First principles-based density functional theory (DFT) calculations also show that there is a first-order transition between tetragonal and collapsed tetragonal (cT) phases, with an onset near ~ 10 GPa as well as the appearance of the ferromagnetic (FM) ordering in the cT phase. Above ~ 30 GPa, the experimental signatures of the magnetic ordering vanish in a first-order-like manner, consistent with the theoretical calculation results, indicating that SrCo2P2 is another example of the avoidance of quantum criticality in ferromagnetic intermetallic compounds. SrCo2P2 provides clear evidence that the structural, electronic and magnetic properties associated with the cT transition are strongly entangled and are not only qualitatively captured by our first principles-based calculations but are quantitatively reproduced as well., Comment: 41 pages, 4 figures in main text, 13 figures in SI, 14 tables in SI

The Legacy Survey of Space and Time (LSST) at Vera C. Rubin Observatory is planned to begin in the Fall of 2025. The LSST survey cadence has been designed via a community-driven process regulated by the Survey Cadence Optimization Committee (SCOC), which recommended up to 3% of the observing time to carry out Target of Opportunity (ToO) observations. Experts from the scientific community, Rubin Observatory personnel, and members of the SCOC were brought together to deliver a recommendation for the implementation of the ToO program during a workshop held in March 2024. Four main science cases were identified: gravitational wave multi-messenger astronomy, high energy neutrinos, Galactic supernovae, and small potentially hazardous asteroids possible impactors. Additional science cases were identified and briefly addressed in the documents, including lensed or poorly localized gamma-ray bursts and twilight discoveries. Trigger prioritization, automated response, and detailed strategies were discussed for each science case. This document represents the outcome of the Rubin ToO 2024 workshop, with additional contributions from members of the Rubin Science Collaborations. The implementation of the selection criteria and strategies presented in this document has been endorsed in the SCOC Phase 3 Recommendations document (PSTN-056). Although the ToO program is still to be finalized, this document serves as a baseline plan for ToO observations with the Rubin Observatory.

The flow over a NACA0012 airfoil at a moderate Reynolds number Re = 50,000 and angle of attack of alpha = 3 degrees is investigated using the lattice-Boltzmann method (LBM). The LBM solutions are computed in direct numerical simulation (DNS) mode, i.e., without a wall model. A validation is performed against a Navier-Stokes wall-resolved large eddy simulation, and good agreement is achieved between the different approaches, showing that the LBM can provide accurate solutions of boundary layers under transitional regime, but with a significant computational cost reduction. A laminar separation bubble (LSB) forms over the suction side of the airfoil, leading to intermittent vortex shedding that impacts transition to turbulence and the generation of strong spanwise-coherent vortices. Different shedding patterns are observed including the advection of single vortical structures and pairing of two vortices, which may or may not break into finer turbulent scales. Such flow features are characterized by 2D and 3D events that directly impact the sound generation by the trailing edge. Frequency and amplitude modulations from the LSB lead to a noise spectrum with a main tone plus equidistant secondary tones, and a time-frequency analysis shows that the main tones may switch frequencies due to intermittency. This research advances in the comprehension of the LSB behavior in transitional airfoil flows, impacting the performance and noise generation of blades and propellers., Comment: 31 pages, 16 figures, preprint submitted to the Aerospace Science and Technology

Top quark production in association with a photon offers a unique test ground for the standard model predictions, as it is sensitive to the top-photon coupling. These processes are rare when compared to standard top pair production, however the large amounts of data delivered by the LHC open the window to precise measurements. This talk covered the recent inclusive and differential measurements of top quark single and pair production in association with a photon, by the ATLAS and CMS Collaborations. Potential modifications to the top-photon couplings with respect to the standard model predictions are also explored using the standard model effective field theory., Comment: 7 pages, 4 figures, Talk at the 17th International Workshop on Top Quark Physics (Top2024), 22-27 September 2024

The human brain encodes stimuli from the environment into representations that form a sensory perception of the world. Despite recent advances in understanding visual and auditory perception, olfactory perception remains an under-explored topic in the machine learning community due to the lack of large-scale datasets annotated with labels of human olfactory perception. In this work, we ask the question of whether pre-trained transformer models of chemical structures encode representations that are aligned with human olfactory perception, i.e., can transformers smell like humans? We demonstrate that representations encoded from transformers pre-trained on general chemical structures are highly aligned with human olfactory perception. We use multiple datasets and different types of perceptual representations to show that the representations encoded by transformer models are able to predict: (i) labels associated with odorants provided by experts; (ii) continuous ratings provided by human participants with respect to pre-defined descriptors; and (iii) similarity ratings between odorants provided by human participants. Finally, we evaluate the extent to which this alignment is associated with physicochemical features of odorants known to be relevant for olfactory decoding., Comment: Spotlight paper at NeurIPS 2024

In this paper, we develop a generalized theory of convolutional signal processing and neural networks for Reproducing Kernel Hilbert Spaces (RKHS). Leveraging the theory of algebraic signal processing (ASP), we show that any RKHS allows the formal definition of multiple algebraic convolutional models. We show that any RKHS induces algebras whose elements determine convolutional operators acting on RKHS elements. This approach allows us to achieve scalable filtering and learning as a byproduct of the convolutional model, and simultaneously take advantage of the well-known benefits of processing information in an RKHS. To emphasize the generality and usefulness of our approach, we show how algebraic RKHS can be used to define convolutional signal models on groups, graphons, and traditional Euclidean signal spaces. Furthermore, using algebraic RKHS models, we build convolutional networks, formally defining the notion of pointwise nonlinearities and deriving explicit expressions for the training. Such derivations are obtained in terms of the algebraic representation of the RKHS. We present a set of numerical experiments on real data in which wireless coverage is predicted from measurements captured by unmaned aerial vehicles. This particular real-life scenario emphasizes the benefits of the convolutional RKHS models in neural networks compared to fully connected and standard convolutional operators.

We explore the impact of introducing IIIA-group element Ga in place of IIB-group element Zn in binary intermetallic GdZn on its magnetic and magnetocaloric properties, as well as explicate the modified electronic band structure of the compound. The magnetic transition temperature of the compound decreases with the increase of Ga concentration in GdZn1-xGax (x = 0-0.1) while the crystal structure (CsCl-prototype) and lattice parameters remain unchanged. Our detailed analysis of magnetization and magnetocaloric data conclusively proves that long-ranged magnetic ordering exists in the sample, despite the magnetic interaction considerably weakening with the increase of Ga. The experimental data is rationalized using both theoretical machine learning model and first-principle density functional theory.The electronic band structure of GdZn is manifested with some unusual complex features which gradually diminish with Ga doping and conventional sinusoidal feature of Ruderman-Kittel-Kasuya- Yosida (RKKY)-type interactions also disappears. A mean-field theory model is developed and can successfully describe the overall magnetocaloric behavior of the GdZn1-xGax series of samples

Educational institutions are essential for economic and social development. Budget cuts in Brazil in recent years have made it difficult to carry out their activities and projects. In the case of expenses with water and electricity, unexpected situations can occur, such as leaks and equipment failures, which make their management challenging. This study proposes a comparison between two machine learning models, Random Forest (RF) and Support Vector Regression (SVR), for water and electricity consumption forecasting at the Federal Institute of Paran\'a-Campus Palmas, with a 12-month forecasting horizon, as well as evaluating the influence of the application of climatic variables as exogenous features. The data were collected over the past five years, combining details pertaining to invoices with exogenous and endogenous variables. The two models had their hyperpa-rameters optimized using the Genetic Algorithm (GA) to select the individuals with the best fitness to perform the forecasting with and without climatic variables. The absolute percentage errors and root mean squared error were used as performance measures to evaluate the forecasting accuracy. The results suggest that in forecasting water and electricity consumption over a 12-step horizon, the Random Forest model exhibited the most superior performance. The integration of climatic variables often led to diminished forecasting accuracy, resulting in higher errors. Both models still had certain difficulties in predicting water consumption, indicating that new studies with different models or variables are welcome., Comment: Conference: International Joint Conference on Industrial Engineering and Operations Management (IJCIEOM ). At: Salvador-BA, Brazil

In this work we improve the dynamic-evolutionary framework of two massive clusters at intermediate redshifts: Cl 0024+17 at $z \sim 0.4$ and MS 0451-03 at $z \sim 0.5$. The spectroscopic galaxy members were selected from Moran et al. (2007a), which combine optical and UV imaging with spectroscopy. Using a set of dynamic estimators with different approaches, our results show that both Cl 0024+17 and MS 0451-03 are non-relaxed systems with distinct dynamical configurations. Cl 0024+17 exhibits a disturbed kinematics, displaying significant gaps and a velocity dispersion profile suggesting a merger. This is confirmed by the presence of previously reported substructures and new ones identified in this study. MS 0451-03 appears less disturbed than Cl 0024+17, indicating by the significant segregation between late and early-type galaxies, with the latter occupying more central regions of the projected phase-space. However, five previously unobserved substructures and non-Gaussianity in the velocity distribution indicate that MS 0451-03 is also out of equilibrium. In both clusters, there are substructures infalling onto the systems, indicating key moments in their assembly histories and potential effects on the pre-processing of galaxies within these subgroups. This is suggested by the high percentage of early-type galaxies outside $R_{200}$ (approximately $83\%$) in the case of CL 0024+17. This work reinforces the importance of more detailed dynamical analysis of clusters to better characterize their evolutionary picture., Comment: Accepted for publication in MNRAS, 16 pages, 11 Figures

A triangular optical cavity can be used to distinguish between two beams with the same intensity profile but different wavefronts. This is what we show in this paper, both theoretically and experimentally, in the case of beams with a doughnut-like intensity profile: one of them having a helical wavefront (vortex beam with orbital angular momentum) and the other with no orbital angular momentum at all (which we call pseudovortex beam). We write the mode decomposition of such beams in the Hermite-Gaussian basis and in the the Laguerre-Gauss basis, respectively, and study how they interact with a triangular cavity in terms of their resonance peaks. The experimental results corroborate the theory, showing that each beam displays its own resonance pattern. Therefore, such cavity may be used to identify beams with orbital angular momentum, distinguishing them from pseudovortices., Comment: 5 pages, 3 figures

Autism Spectrum Disorder (ASD) is a neurodivergent condition with a wide range of characteristics and support levels. Individuals with ASD can exhibit various combinations of traits such as difficulties in social interaction, communication, and language, alongside restricted interests and repetitive activities. Many adults with ASD live independently due to increased awareness and late diagnoses, which help them manage long-standing challenges. Predictability, clarity, and minimized sensory stimuli are crucial for the daily comfort of autistic individuals. In mobile applications, autistic users face significant cognitive overload compared to neurotypicals, resulting in higher effort and time to complete tasks. Urban mobility apps, essential for daily routines, often overlook the needs of autistic users, leading to cognitive overload issues. This study investigates the accessibility of urban mobility apps for autistic individuals using the Interfaces Accessibility Guide for Autism (GAIA). By evaluating various apps, we have identified a common gap regarding accessibility for people with Autism Spectrum Disorder (ASD). This limitation relates to the absence of a functionality that allows users on the autism spectrum to customize the characteristics of the textual and visual elements of the software, such as changing the text font, altering the font type, and adjusting text colors, as well as native audio guidance within the applications themselves. Currently, the only function in this context is for visually impaired people, which completely changes the user experience in terms of navigation., Comment: Author pre-print of article published in the XXIII Brazilian Symposium on Human Factors in Computing Systems (IHC'24)

In this paper, we introduce the concept of a nested family of torsion pairs and will prove that this concept is strongly related to the existence of stratifying systems. Specifically, every stratifying system induces a nested family of torsion pairs. Moreover, every stratifying system can be obtained as a quotient or submodule of a module that admits a certain direct sum decomposition with respect to the nested family of torsion pairs. Additionally, we present a stratifying system of infinite size that cannot be indexed by $\mathbb{N}$., Comment: 24 pages

Adversarial training can be used to learn models that are robust against perturbations. For linear models, it can be formulated as a convex optimization problem. Compared to methods proposed in the context of deep learning, leveraging the optimization structure allows significantly faster convergence rates. Still, the use of generic convex solvers can be inefficient for large-scale problems. Here, we propose tailored optimization algorithms for the adversarial training of linear models, which render large-scale regression and classification problems more tractable. For regression problems, we propose a family of solvers based on iterative ridge regression and, for classification, a family of solvers based on projected gradient descent. The methods are based on extended variable reformulations of the original problem. We illustrate their efficiency in numerical examples.

Owing to their similarities with giant exoplanets, brown dwarf companions of stars provide insights into the fundamental processes of planet formation and evolution. From their orbits, several brown dwarf companions are found to be more massive than theoretical predictions given their luminosities and the ages of their host stars (e.g. Brandt et al. 2021, Cheetham et al. 2018, Li et al. 2023). Either the theory is incomplete or these objects are not single entities. For example, they could be two brown dwarfs each with a lower mass and intrinsic luminosity (Brandt et al. 2021, Howe et al. 2024). The most problematic example is Gliese 229 B (Nakajima et al. 1995, Oppenheimer et al. 1995), which is at least 2-6 times less luminous than model predictions given its dynamical mass of $71.4\pm0.6$ Jupiter masses ($M_{\rm Jup}$) (Brandt et al. 2021). We observed Gliese 229 B with the GRAVITY interferometer and, separately, the CRIRES+ spectrograph at the Very Large Telescope. Both sets of observations independently resolve Gliese 229 B into two components, Gliese 229 Ba and Bb, settling the conflict between theory and observations. The two objects have a flux ratio of $0.47\pm0.03$ at a wavelength of 2 $\mu$m and masses of $38.1\pm1.0$ and $34.4\pm1.5$ $M_{\rm Jup}$, respectively. They orbit each other every 12.1 days with a semimajor axis of 0.042 astronomical units (AU). The discovery of Gliese 229 BaBb, each only a few times more massive than the most massive planets, and separated by 16 times the Earth-moon distance, raises new questions about the formation and prevalence of tight binary brown dwarfs around stars., Comment: Published in Nature. The Version of Record of this article is located at https://www.nature.com/articles/s41586-024-08064-x

Inspired by recent visible pump - infrared probe spectra reported for molecular catalysts adsorbed to quantum dots, we introduce a theory of non-equilibrium vibronic Fano resonances arising from the interference of quantum dot excited-state intraband transitions and infrared vibrational excitations of a molecular adsorbate. Our theory suggests a superexchange mechanism for the observed Fano resonances where charge-transfer states mediate the effective interaction between molecular vibrations and near-resonance quantum dot intraband transitions. We present a perturbative treatment of the effective adsorbate-quantum dot vibronic interaction and employ it to construct a two-reservoir Fano model that enables us to capture key experimental trends, including the relationship between the Fano asymmetry factor, quantum dot size, and the distance between the molecular adsorbate and the quantum dot. We focus on adsorbate-quantum dot species, but our theory's implications are shown to be generic for molecules adsorbed to materials with resonant strong mid-infrared electronic transitions., Comment: 10 pages, 7 figures

The recent synthesis of nitrogen-doped monolayer amorphous carbon (MAC @N) opens new possibilities for multifunctional materials. In this study, we have investigated the nitrogen doping limits and their effects on MAC@N's structural and electronic properties using density functional-based tight-binding simulations. Our results show that MAC@N remains stable up to 35\% nitrogen doping, beyond which the lattice becomes unstable. The formation energies of MAC@N are higher than those of nitrogen-doped graphene for all the cases we have investigated. Both undoped MAC and MAC@N exhibit metallic behavior, although only MAC features a Dirac-like cone. MAC has an estimated Young's modulus value of about 410 GPa, while MAC@N's modulus can vary around 416 GPa depending on nitrogen content. MAC displays optical activity in the ultraviolet range, whereas MAC@N features light absorption within the infrared and visible ranges, suggesting potential for distinct optoelectronic applications. Their structural thermal stabilities were addressed through molecular dynamics simulations. MAC melts at approximately 4900K, while MAC@N loses its structural integrity for temperatures ranging from 300K to 3300K, lower than graphene. These results point to potential MAC@N applications in flexible electronics and optoelectronics., Comment: 15 pages

Existing retrieval-based reasoning approaches for large language models (LLMs) heavily rely on the density and quality of the non-parametric knowledge source to provide domain knowledge and explicit reasoning chain. However, inclusive knowledge sources are expensive and sometimes infeasible to build for scientific or corner domains. To tackle the challenges, we introduce Graph Inspired Veracity Extrapolation (GIVE), a novel reasoning framework that integrates the parametric and non-parametric memories to enhance both knowledge retrieval and faithful reasoning processes on very sparse knowledge graphs. By leveraging the external structured knowledge to inspire LLM to model the interconnections among relevant concepts, our method facilitates a more logical and step-wise reasoning approach akin to experts' problem-solving, rather than gold answer retrieval. Specifically, the framework prompts LLMs to decompose the query into crucial concepts and attributes, construct entity groups with relevant entities, and build an augmented reasoning chain by probing potential relationships among node pairs across these entity groups. Our method incorporates both factual and extrapolated linkages to enable comprehensive understanding and response generation. Extensive experiments on reasoning-intense benchmarks on biomedical and commonsense QA demonstrate the effectiveness of our proposed method. Specifically, GIVE enables GPT3.5-turbo to outperform advanced models like GPT4 without any additional training cost, thereby underscoring the efficacy of integrating structured information and internal reasoning ability of LLMs for tackling specialized tasks with limited external resources.

Fej\'er monotonicity is a well-established property commonly observed in sequences generated by optimization algorithms. In this paper, we introduce an extension of this property, called Fej\'er* monotonicity, which was initially proposed in [SIAM J. Optim., 34(3), 2535-2556 (2024)]. We discuss and build upon the concept by exploring its behavior within Hilbert spaces, presenting an illustrative example and insightful results regarding weak and strong convergence. We also compare Fej\'er* monotonicity with other weak notions of Fej\'er-like monotonicity, to better establish the role of Fej\'er* monotonicity in optimization algorithms.

Observational evidence consistently shows that the universe is spatially flat and undergoes Lorentzian time dilation as a function of redshift. In combination, such discoveries suggest that a Minkowskian description of cosmology might be technically viable. The thermal evolution that transpires in a conformal spacetime is herein derived. The description is constrained by the energy conservation of a unified cosmic fluid. The resulting model puts forth a Lorentzian correction for the temperature of the CMB as a function of redshift, which improves current data fitting without adding any free parameter. Furthermore, it sheds light upon the early galaxy formation problem: our model predicts up to 0.86 Gyr older objects within the first two billion years of the structure evolution in the universe., Comment: 14 pages, 8 figures

It is generally assumed that the entropy of activation associated with dislocation glide, a process controlling body-centered cubic metals strength, is either constant or related to the enthalpy and stress through the empirical Meyer-Neldel (MN) law. We present an atomistic investigation revealing that while both cases can be encountered depending on the refinement of simulation methodology, the MN law for this process is an artifact of simple radial models of interatomic interaction which predict a rugged energy landscape, and is not necessarily connected to the temperature-evolution of the bulk modulus as historically thought. Comparing a classical and a dedicated machine-learning (ML) interatomic potential, we show surprisingly large differences between the classical harmonic theory and anharmonic calculations with the classical model, impacting dislocation velocities by orders of magnitude, while both predictions are in excellent agreement using the ML, which predicts a constant activation entropy. Simulations based on classical potentials, a cornerstone of materials modelling, could be broadly affected by these over-estimated entropic effects.

We investigate the prominent class of fair representation learning methods for bias mitigation. Using causal reasoning to define and formalise different sources of dataset bias, we reveal important implicit assumptions inherent to these methods. We prove fundamental limitations on fair representation learning when evaluation data is drawn from the same distribution as training data and run experiments across a range of medical modalities to examine the performance of fair representation learning under distribution shifts. Our results explain apparent contradictions in the existing literature and reveal how rarely considered causal and statistical aspects of the underlying data affect the validity of fair representation learning. We raise doubts about current evaluation practices and the applicability of fair representation learning methods in performance-sensitive settings. We argue that fine-grained analysis of dataset biases should play a key role in the field moving forward.

Low Rank Adaptation (LoRA) is a popular Parameter Efficient Fine Tuning (PEFT) method that effectively adapts large pre-trained models for downstream tasks. LoRA parameterizes model updates using low-rank matrices at each layer, significantly reducing the number of trainable parameters and, consequently, resource requirements during fine-tuning. However, the lower bound on the number of trainable parameters remains high due to the use of the low-rank matrix model. In this paper, we address this limitation by proposing a novel approach that employs a low rank tensor parametrization for model updates. The proposed low rank tensor model can significantly reduce the number of trainable parameters, while also allowing for finer-grained control over adapter size. Our experiments on Natural Language Understanding, Instruction Tuning, Preference Optimization and Protein Folding benchmarks demonstrate that our method is both efficient and effective for fine-tuning large language models, achieving a substantial reduction in the number of parameters while maintaining comparable performance.

Network Slicing (NS) has transformed the landscape of resource sharing in networks, offering flexibility to support services and applications with highly variable requirements in areas such as the next-generation 5G/6G mobile networks (NGMN), vehicular networks, industrial Internet of Things (IoT), and verticals. Although significant research and experimentation have driven the development of network slicing, existing architectures often fall short in intrinsic architectural intelligent security capabilities. This paper proposes an architecture-intelligent security mechanism to improve the NS solutions. We idealized a security-native architecture that deploys intelligent microservices as federated agents based on machine learning, providing intra-slice and architectural operation security for the Slicing Future Internet Infrastructures (SFI2) reference architecture. It is noteworthy that federated learning approaches match the highly distributed modern microservice-based architectures, thus providing a unifying and scalable design choice for NS platforms addressing both service and security. Using ML-Agents and Security Agents, our approach identified Distributed Denial-of-Service (DDoS) and intrusion attacks within the slice using generic and non-intrusive telemetry records, achieving an average accuracy of approximately $95.60\%$ in the network slicing architecture and $99.99\%$ for the deployed slice -- intra-slice. This result demonstrates the potential for leveraging architectural operational security and introduces a promising new research direction for network slicing architectures., Comment: 18 pages, 12 figures, Future Generation Computer Systems (FGCS)

The proliferation of digital microscopy images, driven by advances in automated whole slide scanning, presents significant opportunities for biomedical research and clinical diagnostics. However, accurately annotating densely packed information in these images remains a major challenge. To address this, we introduce DiffKillR, a novel framework that reframes cell annotation as the combination of archetype matching and image registration tasks. DiffKillR employs two complementary neural networks: one that learns a diffeomorphism-invariant feature space for robust cell matching and another that computes the precise warping field between cells for annotation mapping. Using a small set of annotated archetypes, DiffKillR efficiently propagates annotations across large microscopy images, reducing the need for extensive manual labeling. More importantly, it is suitable for any type of pixel-level annotation. We will discuss the theoretical properties of DiffKillR and validate it on three microscopy tasks, demonstrating its advantages over existing supervised, semi-supervised, and unsupervised methods.

Generative AI, particularly Language Models (LMs), has the potential to transform real-world domains with societal impact, particularly where access to experts is limited. For example, in education, training novice educators with expert guidance is important for effectiveness but expensive, creating significant barriers to improving education quality at scale. This challenge disproportionately harms students from under-served communities, who stand to gain the most from high-quality education. We introduce Tutor CoPilot, a novel Human-AI approach that leverages a model of expert thinking to provide expert-like guidance to tutors as they tutor. This study is the first randomized controlled trial of a Human-AI system in live tutoring, involving 900 tutors and 1,800 K-12 students from historically under-served communities. Following a preregistered analysis plan, we find that students working with tutors that have access to Tutor CoPilot are 4 percentage points (p.p.) more likely to master topics (p<0.01). Notably, students of lower-rated tutors experienced the greatest benefit, improving mastery by 9 p.p. We find that Tutor CoPilot costs only $20 per-tutor annually. We analyze 550,000+ messages using classifiers to identify pedagogical strategies, and find that tutors with access to Tutor CoPilot are more likely to use high-quality strategies to foster student understanding (e.g., asking guiding questions) and less likely to give away the answer to the student. Tutor interviews highlight how Tutor CoPilot's guidance helps tutors to respond to student needs, though they flag issues in Tutor CoPilot, such as generating suggestions that are not grade-level appropriate. Altogether, our study of Tutor CoPilot demonstrates how Human-AI systems can scale expertise in real-world domains, bridge gaps in skills and create a future where high-quality education is accessible to all students., Comment: Our pre-registration for this randomized controlled trial can be found here: https://osf.io/8d6ha

The energy gap of Bernal-stacked bilayer graphene can be tuned by applying a perpendicular electric field. The origin of this gap can be traced down to the breaking of its inversion symmetry by an onsite potential difference between the layers. This degree of tunability makes bilayer graphene a perfect playground for the study of the effects of electric fields, such as the crystalline field, which are developed when layers of other materials are deposited on top of it. Here, we introduce a novel device architecture allowing a simultaneous control over the applied displacement field and the crystalline alignment between two materials. Our experimental and numerical results confirm that the crystal field and electrostatic doping due to the interface reflect the 120$^{\circ}$ symmetry of the bilayer graphene/BN heterostructure and are highly affected by the commensurate state. These results provide an unique insight into the role of the twist angle in the development of internal crystal fields and intrinsic electrostatic doping in heterostructures. Our results highligth the importance of layer alignment, beyond the existence of a moir\'e superlattice, to understand and simulate the intrinsic properties of a van der Waal heterostructure.

Accurate sleep stage classification is crucial for diagnosing sleep disorders and evaluating sleep quality. While polysomnography (PSG) remains the gold standard, photoplethysmography (PPG) is more practical due to its affordability and widespread use in wearable devices. However, state-of-the-art sleep staging methods often require prolonged continuous signal acquisition, making them impractical for wearable devices due to high energy consumption. Shorter signal acquisitions are more feasible but less accurate. Our work proposes an adapted sleep staging model based on top-performing state-of-the-art methods and evaluates its performance with different PPG segment sizes. We concatenate 30-second PPG segments over 15-minute intervals to leverage longer segment contexts. This approach achieved an accuracy of 0.75, a Cohen's Kappa of 0.60, an F1-Weighted score of 0.74, and an F1-Macro score of 0.60. Although reducing segment size decreased sensitivity for deep and REM stages, our strategy outperformed single 30-second window methods, particularly for these stages., Comment: 11 pages, 5 figures, 1 table

The goal of continual learning (CL) is to train a model that can solve multiple tasks presented sequentially. Recent CL approaches have achieved strong performance by leveraging large pre-trained models that generalize well to downstream tasks. However, such methods lack theoretical guarantees, making them prone to unexpected failures. Conversely, principled CL approaches often fail to achieve competitive performance. In this work, we bridge this gap between theory and practice by integrating an empirically strong approach (RanPAC) into a principled framework, Ideal Continual Learner (ICL), designed to prevent forgetting. Specifically, we lift pre-trained features into a higher dimensional space and formulate an over-parametrized minimum-norm least-squares problem. We find that the lifted features are highly ill-conditioned, potentially leading to large training errors (numerical instability) and increased generalization errors (double descent). We address these challenges by continually truncating the singular value decomposition (SVD) of the lifted features. Our approach, termed ICL-TSVD, is stable with respect to the choice of hyperparameters, can handle hundreds of tasks, and outperforms state-of-the-art CL methods on multiple datasets. Importantly, our method satisfies a recurrence relation throughout its continual learning process, which allows us to prove it maintains small training and generalization errors by appropriately truncating a fraction of SVD factors. This results in a stable continual learning method with strong empirical performance and theoretical guarantees., Comment: 45 pages, 19 figures, 14 tables (Preprint, Oct 1, 2024)

Neurological disorders that affect speech production, such as Alzheimer's Disease (AD), significantly impact the lives of both patients and caregivers, whether through social, psycho-emotional effects or other aspects not yet fully understood. Recent advancements in Large Language Model (LLM) architectures have developed many tools to identify representative features of neurological disorders through spontaneous speech. However, LLMs typically lack interpretability, meaning they do not provide clear and specific reasons for their decisions. Therefore, there is a need for methods capable of identifying the representative features of neurological disorders in speech and explaining clearly why these features are relevant. This paper presents an explainable LLM method, named SLIME (Statistical and Linguistic Insights for Model Explanation), capable of identifying lexical components representative of AD and indicating which components are most important for the LLM's decision. In developing this method, we used an English-language dataset consisting of transcriptions from the Cookie Theft picture description task. The LLM Bidirectional Encoder Representations from Transformers (BERT) classified the textual descriptions as either AD or control groups. To identify representative lexical features and determine which are most relevant to the model's decision, we used a pipeline involving Integrated Gradients (IG), Linguistic Inquiry and Word Count (LIWC), and statistical analysis. Our method demonstrates that BERT leverages lexical components that reflect a reduction in social references in AD and identifies which further improve the LLM's accuracy. Thus, we provide an explainability tool that enhances confidence in applying LLMs to neurological clinical contexts, particularly in the study of neurodegeneration., Comment: 27 pages, 6 figures, authors Marina Ribeiro and B\'arbara Malcorra have equal contribution, C\'esar Renn\'o-Costa is the corresponding author

This study evaluates the performance of large language models (LLMs) as medical agents in Portuguese, aiming to develop a reliable and relevant virtual assistant for healthcare professionals. The HealthCareMagic-100k-en and MedQuAD datasets, translated from English using GPT-3.5, were used to fine-tune the ChatBode-7B model using the PEFT-QLoRA method. The InternLM2 model, with initial training on medical data, presented the best overall performance, with high precision and adequacy in metrics such as accuracy, completeness and safety. However, DrBode models, derived from ChatBode, exhibited a phenomenon of catastrophic forgetting of acquired medical knowledge. Despite this, these models performed frequently or even better in aspects such as grammaticality and coherence. A significant challenge was low inter-rater agreement, highlighting the need for more robust assessment protocols. This work paves the way for future research, such as evaluating multilingual models specific to the medical field, improving the quality of training data, and developing more consistent evaluation methodologies for the medical field., Comment: This work has been submitted to the IEEE for possible publication

The widespread use of machine learning in credit scoring has brought significant advancements in risk assessment and decision-making. However, it has also raised concerns about potential biases, discrimination, and lack of transparency in these automated systems. This tutorial paper performed a non-systematic literature review to guide best practices for developing responsible machine learning models in credit scoring, focusing on fairness, reject inference, and explainability. We discuss definitions, metrics, and techniques for mitigating biases and ensuring equitable outcomes across different groups. Additionally, we address the issue of limited data representativeness by exploring reject inference methods that incorporate information from rejected loan applications. Finally, we emphasize the importance of transparency and explainability in credit models, discussing techniques that provide insights into the decision-making process and enable individuals to understand and potentially improve their creditworthiness. By adopting these best practices, financial institutions can harness the power of machine learning while upholding ethical and responsible lending practices.

Unlabeled motion planning involves assigning a set of robots to target locations while ensuring collision avoidance, aiming to minimize the total distance traveled. The problem forms an essential building block for multi-robot systems in applications such as exploration, surveillance, and transportation. We address this problem in a decentralized setting where each robot knows only the positions of its $k$-nearest robots and $k$-nearest targets. This scenario combines elements of combinatorial assignment and continuous-space motion planning, posing significant scalability challenges for traditional centralized approaches. To overcome these challenges, we propose a decentralized policy learned via a Graph Neural Network (GNN). The GNN enables robots to determine (1) what information to communicate to neighbors and (2) how to integrate received information with local observations for decision-making. We train the GNN using imitation learning with the centralized Hungarian algorithm as the expert policy, and further fine-tune it using reinforcement learning to avoid collisions and enhance performance. Extensive empirical evaluations demonstrate the scalability and effectiveness of our approach. The GNN policy trained on 100 robots generalizes to scenarios with up to 500 robots, outperforming state-of-the-art solutions by 8.6\% on average and significantly surpassing greedy decentralized methods. This work lays the foundation for solving multi-robot coordination problems in settings where scalability is important., Comment: 6 pages, 6 figures, submitted to ICRA 2025