Search

Context. Spectroscopic surveys like APOGEE, GALAH, and LAMOST have significantly advanced our understanding of the Milky Way by providing extensive stellar parameters and chemical abundances. Complementing these, photometric surveys with narrow/medium-band filters, such as the Southern Photometric Local Universe Survey (S-PLUS), offer the potential to estimate stellar parameters and abundances for a much larger number of stars. Aims. This work develops methodologies to extract stellar atmospheric parameters and selected chemical abundances from S-PLUS photometric data, which spans ~3000 square degrees using seven narrowband and five broadband filters. Methods. Using 66 S-PLUS colors, we estimated parameters based on training samples from LAMOST, APOGEE, and GALAH, applying Cost-Sensitive Neural Networks (NN) and Random Forests (RF). We tested for spurious correlations by including abundances not covered by the S-PLUS filters and evaluated NN and RF performance, with NN consistently outperforming RF. Including Teff and log g as features improved accuracy by ~3%. We retained only parameters with a goodness-of-fit above 50%. Results. Our approach provides reliable estimates of fundamental parameters (Teff, log g, [Fe/H]) and abundance ratios such as [{\alpha}/Fe], [Al/Fe], [C/Fe], [Li/Fe], and [Mg/Fe] for ~5 million stars, with goodness-of-fit >60%. Additional ratios like [Cu/Fe], [O/Fe], and [Si/Fe] were derived but are less accurate. Validation using star clusters, TESS, and J-PLUS data confirmed the robustness of our methodology. Conclusions. By leveraging S-PLUS photometry and machine learning, we present a cost-effective alternative to high-resolution spectroscopy for deriving stellar parameters and abundances, enabling insights into Milky Way stellar populations and supporting future classification efforts., Comment: 23 pages, 14 Figures

We surface a new threat to closed-weight Large Language Models (LLMs) that enables an attacker to compute optimization-based prompt injections. Specifically, we characterize how an attacker can leverage the loss-like information returned from the remote fine-tuning interface to guide the search for adversarial prompts. The fine-tuning interface is hosted by an LLM vendor and allows developers to fine-tune LLMs for their tasks, thus providing utility, but also exposes enough information for an attacker to compute adversarial prompts. Through an experimental analysis, we characterize the loss-like values returned by the Gemini fine-tuning API and demonstrate that they provide a useful signal for discrete optimization of adversarial prompts using a greedy search algorithm. Using the PurpleLlama prompt injection benchmark, we demonstrate attack success rates between 65% and 82% on Google's Gemini family of LLMs. These attacks exploit the classic utility-security tradeoff - the fine-tuning interface provides a useful feature for developers but also exposes the LLMs to powerful attacks.

In this work, we revisited the Ziff-Gullari-Barshad (ZGB) model in order to investigate its critical behavior when carbon monoxide (CO) molecules are allowed to desorb from the catalytic surface. As shown by several authors, when this kind of desorption takes place, the first-order phase transition of the standard model disappears, and an Ising-like critical point is found for a very small value of the desorption rate. However, our time-dependent Monte Carlo simulations reveal that, instead of a single critical point, there exists a critical line that encompasses multiple universality classes, passing through the three- and four-state Potts points, as well as, the Ising one, resulting in an unprecedented critical line of weak universality., Comment: 8 pages, 7 figures

In this paper, we continue the quest to understand the interplay between wrapped Floer homology barcode and topological entropy. Wrapped Floer homology barcode entropy is defined as the exponential growth, with respect to the left endpoints, of the number of not-too-short bars in its barcode. We prove that, in the presence of a topologically transitive, locally maximal hyperbolic set for the Reeb flow on the boundary of a Liouville domain, the barcode entropy is bounded from below by the topological entropy restricted to the hyperbolic set., Comment: 23 pages

Glucose oxidation is a key reaction in biosensing and energy conversion, with anodized copper electrodes showing great potential as catalysts for its enhancement. This study explores the relationship between anodization parameters and glucose oxidation on thin copper films, providing valuable insights for optimization. By adjusting anodization parameters, specific species such as Cu2O, Cu(OH)2, and CuO can be selectively formed, each exhibiting different catalytic activities. Applying polarization in 1 M KOH at 0 V (vs Ag|AgCl) generates a highly active CuO surface layer, leading to significant performance improvements over bare copper electrodes. Specifically, sensitivity increases by 55% in the 0.1 mM-0.5 mM range and 73% in the 0.75 mM-2 mM range. Additionally, the electrodes demonstrate an exceptionally low detection limit of just 0.004 mM. These results are attributed to scalable, reproducible manufacturing on cleanroom-compatible platforms. This study not only sheds light on the effects of thin copper film anodization on glucose oxidation but also provides a practical approach for enhancing the efficiency of copper-based systems in biosensing and energy conversion applications.

Incoherent Diffraction Imaging - IDI - is a diffraction-based imaging technique that has been recently proposed to exploit the partial coherence of incoherently scattered light to retrieve structural information from the scattering centers. Similar to the stellar intensity interferometry of Hanbury Brown and Twiss, the signal builds up on the second-order spatial correlations of the emitted light. The complex spatial distribution of the target is thereby encoded in the spatial intensity fluctuations of the scattered light. The first experimental realisations of this imaging technique have been realised using the fluorescence excited by an ultra-short X-ray pulse at Free Electron Laser (FEL) facilities. Here, we propose an alternative set-up based on a table-top Pseudo-Thermal Light Source. This set-up allows us to explore IDI under a wide range of physically relevant conditions as well as to benchmark numerical and analytical models currently used to determine the imaging capabilities of this technique.

Music genres are shaped by both the stylistic features of songs and the cultural preferences of artists' audiences. Automatic classification of music genres using lyrics can be useful in several applications such as recommendation systems, playlist creation, and library organization. We present a multi-label, cross-lingual genre classification system based on multilingual sentence embeddings generated by sBERT. Using a bilingual Portuguese-English dataset with eight overlapping genres, we demonstrate the system's ability to train on lyrics in one language and predict genres in another. Our approach outperforms the baseline approach of translating lyrics and using a bag-of-words representation, improving the genrewise average F1-Score from 0.35 to 0.69. The classifier uses a one-vs-all architecture, enabling it to assign multiple genre labels to a single lyric. Experimental results reveal that dataset centralization notably improves cross-lingual performance. This approach offers a scalable solution for genre classification across underrepresented languages and cultural domains, advancing the capabilities of music information retrieval systems., Comment: 5 pages

Systems such as Microsoft Songsmith automatically assign chords and harmony to a melody by minimizing the dissonance across all chord changes. Although this produces harmonious music, it is not what practicing musicians do. In this paper, I describe Harmonizer, a prototype system for melodic harmonization. Harmonizer uses Schoenberg's chart of regions as the underlying data structure that allows harmonization using several different methods. Because the chart reveals inter-chordal relationships, the harmonizations may be programmed to emphasize desired relationships. In the prototype Harmonizer, I also explore recent signal-processing methods that enable songwriters to easily input a melody by singing or by playing a musical instrument. The prototype Harmonizer is available on GitHub and a video demonstrating its distinctive harmonizations is on YouTube as explained in the Results section of the paper.

In this paper, we consider the monoids of all endomorphisms, of all weak endomorphisms, of all strong endomorphisms and of all strong weak endomorphisms of a star graph with a finite number of vertices. Our main objective is to exhibit a presentation for each of them., Comment: math.RA. arXiv admin note: text overlap with arXiv:2410.21084

We generalize a recent ``AdS S-matrix" formulation for interacting massive scalars on AdS spacetimes to the case of massive vector fields. This method relies on taking the infinite radius limit for scattering processes perturbatively, which is analyzed using Witten diagrams in the momentum space formulation of global AdS with embedding space coordinates. It recovers the S-matrix with subleading corrections in powers of the inverse AdS radius about a flat spacetime region within the bulk. We first derive the massive vector bulk-to-boundary and bulk-to-bulk propagators within this perturbation theory. As an example, we consider the Abelian Higgs Model in a certain regime of the coupling parameter space to model an interacting Proca theory on AdS spacetimes. We specifically compute the AdS S-matrix for a process involving massive external vector fields mediated by a massive scalar. We lastly discuss possible massless limit of propagators within this perturbative framework., Comment: 35 pages, 4 figures

We investigate the ultrafast structural dynamics of LaAlO3 thin films driven by short mid-infrared laser pulses at 20 THz. Time-resolved X-ray diffraction reveals an immediate lattice expansion and an acoustic breathing mode of the film. First-principles theory and a spring-mass model identify the direct coupling between coherently driven infrared-active phonons and strain as the underlying mechanism. Time-resolved optical birefringence measurements confirm that the amplitude of this acoustic mode scales linearly with the pump fluence, which agrees with the theory. Furthermore, time-resolved X-ray diffuse scattering indicates that THz excitation enhances crystallinity by inducing a non-thermal increase in structural symmetry originating from preexisting defects. These findings highlight the potential of a multimodal approach-combining time-resolved X-ray and optical measurements and first-principles theory-to elucidate and control structural dynamics in nanoscale materials.

EasyVis2 is a system designed for hands-free, real-time 3D visualization during laparoscopic surgery. It incorporates a surgical trocar equipped with a set of micro-cameras, which are inserted into the body cavity to provide an expanded field of view and a 3D perspective of the surgical procedure. A sophisticated deep neural network algorithm, YOLOv8-Pose, is tailored to estimate the position and orientation of surgical instruments in each individual camera view. Subsequently, 3D surgical tool pose estimation is performed using associated 2D key points across multiple views. This enables the rendering of a 3D surface model of the surgical tools overlaid on the observed background scene for real-time visualization. In this study, we explain the process of developing a training dataset for new surgical tools to customize YoLOv8-Pose while minimizing labeling efforts. Extensive experiments were conducted to compare EasyVis2 with the original EasyVis, revealing that, with the same number of cameras, the new system improves 3D reconstruction accuracy and reduces computation time. Additionally, experiments with 3D rendering on real animal tissue visually demonstrated the distance between surgical tools and tissues by displaying virtual side views, indicating potential applications in real surgeries in the future., Comment: 11 pages (12 pages with citations), 11 figures

Brain metastases are a common diagnosis that affects between 20% and 40% of cancer patients. Subsequent to radiation therapy, patients with brain metastases undergo follow-up sessions during which the response to treatment is monitored. In this study, a dataset of medical images from 44 patients with at least one brain metastasis and different primary tumor locations was collected and processed. Each patient was treated with either a linear accelerator or a gamma knife. Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) scans were collected at various time points, including before treatment and during follow-up sessions. The CT datasets were processed using windowing and artifact reduction techniques, while the MRI datasets were subjected to CLAHE. The NifTI files corresponding to the CT and MRI images were made public available. In order to align the datasets of each patient, a multimodal registration was performed between the CT and MRI datasets, with different software options being tested. The fusion matrices were provided together with the dataset. The aforementioned steps resulted in the creation of an optimized dataset, prepared for use in a range of studies related to brain metastases. RFUds is publicity available at zenodo under the DOI 10.5281/zenodo.14524784.

For each positive integer $n$, let $\mathbb F_{q^n}$ be the unique $n$-degree extension of the finite field $\mathbb F_q$ with $q$ elements, where $q$ is a prime power. It is known that for arbitrary $q$ and $n$, there exists an element $\beta\in \mathbb F_{q^n}$ such that its Galois conjugates $\beta, \beta^q, \ldots, \beta^{q^{n-1}}$ form a basis for $\mathbb F_{q^n}$ as an $\mathbb F_q$-vector space. These elements are called normal and they work as additive generators of finite fields. On the other hand, the multiplicative group $\mathbb F_{q^n}^*$ is cyclic and any generator of this group is a primitive element. Many past works have dealt with the existence of primitive and normal elements with specified properties, including the existence of primitive elements whose traces over intermediate extensions are prescribed. Inspired by the latter, in this paper we explore the existence of normal elements whose norms over intermediate extensions are prescribed. We combine combinatorial and number-theoretic ideas and obtain both asymptotic and concrete results. In particular, we completely solve the problem in the case where only one intermediate extension is considered.

We study the phase ordering dynamics of the classical antiferromagnetic $J_1$-$J_2$ (nearest-neighbor and next-nearest-neighbor couplings) Heisenberg model on the square lattice in the strong frustration regime ($J_2/J_1 > 1/2$). While thermal fluctuations preclude any long-range magnetic order at finite temperatures, the system exhibits a long-range spin-driven nematic phase at low temperatures. The transition into the nematic phase is further shown to belong to the two-dimensional Ising universality class based on the critical exponents near the phase transition. Our large-scale stochastic Landau-Lifshitz-Gilbert simulations find a two-stage phase ordering when the system is quenched from a high-temperature paramagnetic state into the nematic phase. In the early stage, collinear alignments of spins lead to a locally saturated Ising-nematic order. Once domains of well-defined Ising order are developed, the late-stage relaxation is dominated by curvature-driven domain coarsening, as described by the Allen-Cahn equation. The characteristic size of Ising-nematic domains scales as the square root of time, similar to the kinetic Ising model described by the time-dependent Ginzburg-Landau theory. Our results confirm that the late-stage ordering kinetics of the spin-driven nematic, which is a vestigial order of the frustrated Heisenberg model, belongs to the dynamical universality class of a non-conserved Ising order. Interestingly, the system shows no violation of the superuniversality hypothesis under weak bond disorder. The dynamic scaling invariance is preserved in the presence of weak bond disorder. We also discuss possible applications of our results to materials for which vestigial Ising-nematic order is realized., Comment: 9 pages, 9 figures

The chemical abundances of elements such as barium and the lanthanides are essential to understand the nucleosynthesis of heavy elements in the early Universe as well as the contribution of different neutron capture processes (for example slow versus rapid) at different epochs. The Chemical Evolution of R-process Elements in Stars (CERES) project aims to provide a homogeneous analysis of a sample of metal-poor stars ( [Fe/H]\<-1.5) to improve our understanding of the nucleosynthesis of neutron capture elements, in particular the r-process elements, in the early Galaxy. Our data consist of a sample of high resolution and high signal-to-noise ratio UVES spectra. The chemical abundances were derived through spectrum synthesis, using the same model atmospheres and stellar parameters as derived in the first paper of the CERES series. We measured chemical abundances or upper limits of seven heavy neutron capture elements (Ba, La, Ce, Pr, Nd, Sm, and Eu) for a sample of 52 metal-poor giant stars. We estimated through the mean shift clustering algorithm that at Ba/H =-2.4 and Fe/H =-2.4 a variation in the trend of X/Ba with X=La,Nd,Sm,Eu, versus Ba/H occurs. This result suggests that, for Ba/H $\<$$-2.4$, Ba nucleosynthesis in the Milky Way halo is primarily due to the $r$-process, while for Ba/H \<-2.4 the effect of the s-process contribution begins to be visible. In our sample, stars with Ba/Eu compatible with a Solar System pure r-process value (hereafter, r-pure) do not show any particular trend compared to other stars, suggesting r-pure stars may form in similar environments to stars with less pure r-process enrichments. Homogeneous investigations of high resolution and signal-to-noise ratio spectra are crucial for studying the heavy elements formation, as they provide abundances that can be used to test nucleosynthesis models as well as Galactic chemical evolution models., Comment: Astronomy & Astrophysics - A&A, In press

Navigating unmanned aerial vehicles in environments where GPS signals are unavailable poses a compelling and intricate challenge. This challenge is further heightened when dealing with Nano Aerial Vehicles (NAVs) due to their compact size, payload restrictions, and computational capabilities. This paper proposes an approach for localization using off-board computing, an off-board monocular camera, and modified open-source algorithms. The proposed method uses three parallel proportional-integral-derivative controllers on the off-board computer to provide velocity corrections via wireless communication, stabilizing the NAV in a custom-controlled environment. Featuring a 3.1cm localization error and a modest setup cost of 50 USD, this approach proves optimal for environments where cost considerations are paramount. It is especially well-suited for applications like teaching drone control in academic institutions, where the specified error margin is deemed acceptable. Various applications are designed to validate the proposed technique, such as landing the NAV on a moving ground vehicle, path planning in a 3D space, and localizing multi-NAVs. The created package is openly available at https://github.com/simmubhangu/eyantra_drone to foster research in this field., Comment: 26 pages. Submitted to Cyber-Physical Systems journal

M dwarfs have become increasingly important in the detection of exoplanets and the study of Earth-sized planets and their habitability. However, 20-30% of M dwarfs have companions that can impact the formation and evolution of planetary systems. We use high-resolution imaging and Gaia astrometry to detect stellar companions around M dwarf exoplanet hosts discovered by TESS and determine the projected separation and estimated stellar masses for each system. We find 47 companions around 216 M dwarfs and a multiplicity rate of $19.4\pm2.7$% that is consistent with field M dwarfs. The binary projected separation distribution is shifted to larger separations, confirming the lack of close binaries hosting transiting exoplanets seen in previous studies. We correct the radii of planets with nearby companions and examine the properties of planets in M dwarf multi-star systems. We also note three multi-planet systems that occur in close binaries ($\lesssim 50$ au) where planet formation is expected to be suppressed., Comment: 28 pages, 13 figures, 5 tables, accepted for publication in The Astronomical Journal

The redshift drift of objects following the cosmological expansion is a unique model-independent probe of background cosmology, detectable by astrophysical facilities presently under construction. Previous forecasts for such measurements assume flat universes. We explore the impact of relaxing this assumption on the constraining power of the redshift drift, focusing on the two most promising routes for its measurement: the SKA at low redshifts, and the Golden Sample for the ELT's ANDES spectrograph at higher redshifts. We also discuss the cosmological sensitivity of possible differential redshift drift measurements, both on their own and, for the specific case of the Golden Sample, in combination with the standard method. Overall, we find that the sensitivity of the redshift drift to curvature is comparable to that of matter (especially at low redshifts) and higher than the sensitivity to the dark energy equation of state. We also show that the sensitivity of redshift drift measurements to these cosmological parameters is asymmetric with respect to the curvature parameter, being different for open and closed universes with the same absolute value of the curvature parameter $\Omega_k$., Comment: 19 pages, 10 figures, 5 tables; Physics of the Dark Universe (in press)

Gravitational wave approximants are widely used tools in gravitational-wave astronomy. They allow for dense coverage of the parameter space of binary black hole (BBH) mergers for purposes of parameter inference, or, more generally, match filtering tasks, while avoiding the computationally expensive full evolution of numerical relativity simulations. However, this comes at a slight cost in terms of accuracy when compared to numerical relativity waveforms, depending on the approach. One way to minimize this is by constructing so-called~\textit{surrogate models} which, instead of using approximate physics or phenomenological formulae, rather interpolate within the space of numerical relativity waveforms. In this work, we introduce~\texttt{NRSurNN3dq4}, a surrogate model for non-precessing BBH merger waveforms powered by neural networks. By relying on the power of deep learning, this approximant is remarkably fast and competitively accurate, as it can generate millions of waveforms in a tenth of a second, while mismatches with numerical relativity waveforms are restrained below $10^{-3}$. We implement this approximant within the~\textsc{bilby} framework for gravitational-wave parameter inference, and show that it it is suitable for parameter estimation tasks.

Effective communication is fundamental to any interaction, yet challenges arise when participants do not share a common language. Automatic translation systems offer a powerful solution to bridge language barriers in such scenarios, but they introduce errors that can lead to misunderstandings and conversation breakdown. A key issue is that current systems fail to incorporate the rich contextual information necessary to resolve ambiguities and omitted details, resulting in literal, inappropriate, or misaligned translations. In this work, we present a framework to improve large language model-based translation systems by incorporating contextual information in bilingual conversational settings. During training, we leverage context-augmented parallel data, which allows the model to generate translations sensitive to conversational history. During inference, we perform quality-aware decoding with context-aware metrics to select the optimal translation from a pool of candidates. We validate both components of our framework on two task-oriented domains: customer chat and user-assistant interaction. Across both settings, our framework consistently results in better translations than state-of-the-art systems like GPT-4o and TowerInstruct, as measured by multiple automatic translation quality metrics on several language pairs. We also show that the resulting model leverages context in an intended and interpretable way, improving consistency between the conveyed message and the generated translations.

The mapping of ocean floor layers is a current challenge for the oil industry. Existing solution methods involve mapping through seismic methods and wave inversion, which are complex and computationally expensive. The introduction of artificial neural networks, specifically UNet, to predict velocity models based on seismic shots reflected from the ocean floor shows promise for optimising this process. In this study, two neural network architectures are validated for velocity model inversion and compared in terms of stability metrics such as loss function and similarity coefficient, as well as the differences between predicted and actual models. Indeed, neural networks prove promising as a solution to this challenge, achieving S{\o}rensen-Dice coefficient values above 70%., Comment: 10 pages, 5 figures. Developed during the 6th Edition of the Advanced School of Experimental Physics (EAFExp), Brazilian Centre for Physics Research

The third r-process peak (Os, Ir, Pt) is poorly understood due to observational challenges, with spectral lines located in the blue or near-ultraviolet region of stellar spectra. These challenges need to be overcome for a better understanding of the r-process in a broader context. To understand how the abundances of the third r-process peak are synthesised and evolve in the Universe, a homogeneous chemical analysis of metal-poor stars using high quality data observed in the blue region of the electromagnetic spectrum (< 400 nm) is necessary. We provide a homogeneous set of abundances for the third r-process peak (Os, Ir, Pt) and Hf, increasing by up to one order of magnitude their availability in the literature. A classical 1D, local thermodynamic equilibrium (LTE) analysis of four elements (Hf, Os, Ir, Pt) is performed, using ATLAS model atmospheres to fit synthetic spectra in high resolution (> 40,000), high signal-to-noise ratio, of 52 red giants observed with UVES/VLT. Due to the heavy line blending involved, a careful determination of upper limits and uncertainties is done. The observational results are compared with state-of-the-art nucleosynthesis models. Our sample displays larger abundances of Ir (Z=77) in comparison to Os (Z=76), which have been measured in a few stars in the past. The results also suggest decoupling between abundances of third r-process peak elements with respect to Eu (rare earth element) in Eu-poor stars. This seems to contradict a co-production scenario of Eu and the third r-process peak elements Os, Ir, and Pt in the progenitors of these objects. Our results are challenging to explain from the nucleosynthetic point of view: the observationally derived abundances indicate the need for an additional early, primary formation channel (or a non-robust r-process)., Comment: 15 pages, 11 figures, Accepted for publication in A&A

Chernyshev, Rauch, and Rautenbach proved that every connected graph on $n$ vertices with less than $\frac{11}{5}n-\frac{18}{5}$ edges has a vertex cut that induces a forest, and conjectured that the same remains true if the graph has less than $3n-6$ edges. We improve their result by proving that every connected graph on $n$ vertices with less than $\frac{9}{4}n$ edges has a vertex cut that induces a forest. We also study weaker versions of the problem that might lead to an improvement on the bound obtained.

Warehouse optimization stands as a critical component for enhancing operational efficiency within the industrial sector. By strategically streamlining warehouse operations, organizations can achieve significant reductions in logistical costs such as the necessary footprint or traveled path, and markedly improve overall workflow efficiency including retrieval times or storage time. Despite the availability of numerous algorithms designed to identify optimal solutions for such optimization challenges, certain scenarios demand computational resources that exceed the capacities of conventional computing systems. In this context, we adapt a formulation of a warehouse optimization problem specifically tailored as a binary optimization problem and implement it in a trapped-ion quantum computer., Comment: 8 pages, 3 figures

We demonstrate sub-Doppler cooling of calcium atoms using a two-photon transition from the ${^1}S_0$ ground state to the upper $4s5s~{^1}S_0$ state via the ${^1}P_1$ intermediate state. We achieve temperatures as low as $260~\mu\text{K}$ in a magneto-optical trap (MOT), well below the Doppler limit ($T_{\text{D}} = 0.8~\text{mK}$) of the ${^1}P_1$ state. We characterize temperature, lifetime and confinement of the MOT over a range of experimental parameters, observing no reduction in lifetime due to coupling to the higher state. We perform theoretical simulations of the cooling scheme and observe good agreement with the experimental results. The two-photon cooling scheme presented in this work provides an alternative to the standard Doppler cooling applied to alkaline-earth atoms, based on a sequence of two magneto-optical traps. The advantages of our scheme are the possibility of varying the effective linewidth of the ${^1}P_1$ state, a higher transfer efficiency (close to 100$\%$), and a more straightforward experimental implementation., Comment: 8 pages 4+1 figures

Research on Reinforcement Learning (RL) approaches for discrete optimization problems has increased considerably, extending RL to an area classically dominated by Operations Research (OR). Vehicle routing problems are a good example of discrete optimization problems with high practical relevance where RL techniques have had considerable success. Despite these advances, open-source development frameworks remain scarce, hampering both the testing of algorithms and the ability to objectively compare results. This ultimately slows down progress in the field and limits the exchange of ideas between the RL and OR communities. Here we propose a library composed of multi-agent environments that simulates classic vehicle routing problems. The library, built on PyTorch, provides a flexible modular architecture design that allows easy customization and incorporation of new routing problems. It follows the Agent Environment Cycle ("AEC") games model and has an intuitive API, enabling rapid adoption and easy integration into existing reinforcement learning frameworks. The library allows for a straightforward use of classical OR benchmark instances in order to narrow the gap between the test beds for algorithm benchmarking used by the RL and OR communities. Additionally, we provide benchmark instance sets for each environment, as well as baseline RL models and training code.

We prove new instances of \emph{Halin's end degree conjecture} within $\mathrm{ZFC}$. In particular, we prove that there is a proper class of cardinals $\kappa$ for which Haliln's end-degree conjecture holds. This answers two questions posed by Geschke, Kurkofka, Melcher and Pitz in 2023. Furthermore, we comment on the relationship between Halin's conjecture and the \emph{Singular Cardinal Hypothesis}, deriving consistency strength from failures of the former. We also show that Halin's conjecture fails on finite intervals of successors of singular cardinals in Meremovich's model., Comment: Issues with text that require major modification

Among the challenges of efficiently managing a factory, inventory management is essential for minimizing operational costs and delivery times. In this paper, we focus on optimizing item allocation in warehouses that use gravity flow racks, which are designed for First In, First Out (FIFO) logistics but present challenges due to the need for frequent item reinsertions during picking operations. We introduce a novel strategy formulated as a QUBO problem, suitable for classical, quantum, and hybrid hardware implementations. By leveraging advances in Adiabatic Quantum Computing and Quantum Annealing, we demonstrate the effectiveness of our strategy through simulations and real-world scenarios. The results highlight the potential of quantum-hybrid approaches to significantly enhance operational efficiency in warehouse management.

Data collection is pervasively bound to our digital lifestyle. A recent study by the IDC reports that the growth of the data created and replicated in 2020 was even higher than in the previous years due to pandemic-related confinements to an astonishing global amount of 64.2 zettabytes of data. While not all the produced data is meant to be analyzed, there are numerous companies whose services/products rely heavily on data analysis. That is to say that mining the produced data has already revealed great value for businesses in different sectors. But to be able to fully realize this value, companies need to be able to hire professionals that are capable of gleaning insights and extracting value from the available data. We hypothesize that people nowadays conducting data-science-related tasks in practice may not have adequate training or formation. So in order to be able to fully support them in being productive in their duties, e.g. by building appropriate tools that increase their productivity, we first need to characterize the current generation of data scientists. To contribute towards this characterization, we conducted a public survey to fully understand who is doing data science, how they work, what are the skills they hold and lack, and which tools they use and need.

Responsible practices for deploying language models include guiding models to recognize and refuse answering prompts that are considered unsafe, while complying with safe prompts. Achieving such behavior typically requires updating model weights, which is costly and inflexible. We explore opportunities to steering model activations at inference time, which does not require updating weights. Using sparse autoencoders, we identify and steer features in Phi-3 Mini that mediate refusal behavior. We find that feature steering can improve Phi-3 Minis robustness to jailbreak attempts across various harms, including challenging multi-turn attacks. However, we discover that feature steering can adversely affect overall performance on benchmarks. These results suggest that identifying steerable mechanisms for refusal via sparse autoencoders is a promising approach for enhancing language model safety, but that more research is needed to mitigate feature steerings adverse effects on performance.

The recent discovery of an axial amplitude (Higgs) mode in the long-studied charge density wave (CDW) systems GdTe$_3$ and LaTe$_3$ suggests a heretofore unidentified hidden order. A theoretical study proposed that the axial Higgs results from a hidden ferroaxial component of the CDW, which could arise from non-trivial orbital texture. Here, we report extensive experimental studies on ErTe$_3$ and HoTe$_3$ that possess a high-temperature CDW similar to other RTe$_3$ (R = rare earth), along with an additional low-temperature CDW with an orthogonal ordering vector. Combining Raman spectroscopy with large-angle convergent beam electron diffraction (LACBED), rotational anisotropy second-harmonic generation (RA-SHG), and muon-spin relaxation ($\mu$SR), we provide unambiguous evidence that the high-temperature CDW breaks translation, rotation, and all vertical and diagonal mirror symmetries, but not time-reversal or inversion. In contrast, the low-temperature CDW only additionally breaks translation symmetry. Simultaneously, Raman scattering shows the high-temperature CDW produces an axial Higgs mode while the low-temperature mode is scalar. The weak monoclinic structural distortion and clear axial response in Raman and SHG are consistent with a ferroaxial phase in RTe$_3$ driven by coupled orbital and charge orders. Thus, our study provides a new standard for uncovering unconventional orders and confirms the power of Higgs modes to reveal them., Comment: 28 pages, 5 figures

We present the discovery of TOI-6303b and TOI-6330b, two massive transiting super-Jupiters orbiting a M0 and a M2 star respectively, as part of the Searching for GEMS survey. These were detected by TESS and then confirmed via ground-based photometry and radial velocity observations with the Habitable-zone Planet Finder (HPF). TOI-6303b has a mass of 7.84 +/- 0.31 MJ, a radius of 1.03 +/- 0.06 RJ , and an orbital period of 9.485 days. TOI-6330b has a mass of 10.00 +/- 0.31 MJ , a radius of 0.97 +/- 0.03 RJ , and an orbital period of 6.850 days. We put these planets in context of super-Jupiters around M-dwarfs discovered from radial-velocity surveys, as well as recent discoveries from astrometry. These planets have masses that can be attributed to two dominant planet formation mechanisms - gravitational instability and core-accretion. Their masses necessitate massive protoplanetary disks that should either be gravitationally unstable, i.e. forming through gravitational instability, or be amongst some of the most massive protoplanetary disks to form objects through core-accretion. We also discuss the eccentricity distribution of these objects, as a potential indicator of their formation and evolutionary mechanisms., Comment: arXiv admin note: text overlap with arXiv:2408.14694

We introduce the one-dimensional Morse-soft-Coulomb (MsC) potential consisting of a Morse repulsive barrier smoothly connected with a soft-core Coulomb potential at the origin. This new potential has a single parameter that controls the softness of the repulsive barrier and the well depth. When this softening-depth parameter tends to zero, the MsC potential approaches the Coulomb potential with an infinite repulsive barrier, a known successful model for the hydrogen atom. We investigate the classical chaotic dynamics of the MsC potential subjected to time-dependent external fields, comparing the results with the Coulomb potential. We show that the MsC potential reproduces the dynamics and the ionization probabilities of the Coulomb potential for sufficiently small values of the softening parameter. We also investigate the role of the softening parameter in the phase-space structures, showing that the increasing of its value leads to the increasing of the chaotic sea and consequently to the rise of the ionization probability. Finally, we address the problem of controlling the dynamics of a particle in the MsC potential from the perspective of optimal control theory, which cannot be easily applied in the case of the Coulomb potential due to the singularity at the origin. We analize a particular optimal solution to the problem of transferring a given amount of energy to the system at minimum cost. Our results show that the MsC potential can be a useful simple model for investigating the hydrogen atom.

Reinforcement learning (RL) has been proven to be an effective and robust method for training neural machine translation systems, especially when paired with powerful reward models that accurately assess translation quality. However, most research has focused on RL methods that use sentence-level feedback, which leads to inefficient learning signals due to the reward sparsity problem -- the model receives a single score for the entire sentence. To address this, we introduce a novel approach that leverages fine-grained token-level reward mechanisms with RL methods. We use xCOMET, a state-of-the-art quality estimation system as our token-level reward model. xCOMET provides detailed feedback by predicting fine-grained error spans and their severity given source-translation pairs. We conduct experiments on small and large translation datasets to compare the impact of sentence-level versus fine-grained reward signals on translation quality. Our results show that training with token-level rewards improves translation quality across language pairs over baselines according to automatic and human evaluation. Furthermore, token-level reward optimization also improves training stability, evidenced by a steady increase in mean rewards over training epochs., Comment: 10 pages, work-in-progress

We prove that if $ T $ is a semi-special tree that is not special, then there exists a graph $ G $, formed as an inflation of a sparse $ T $-graph, such that for any special tree $ S $, $ G $ is not a subdivision of an inflation of an sparse $ S $-graph. Furthermore $G$ has an end of uncountable degree that has no ray graph. This result provides a consistent negative answer to a problem posed by Stefan Geschke et al. in 2023. Additionally, we introduce and explore a property that generalizes Halin's grid theorem, extending it to ends of degree $ \aleph_1 $, which was originally established for ends of countable degree.

Context. Carbon, nitrogen, and oxygen are the most abundant elements throughout the universe, after hydrogen and helium. Studying these elements in low-metallicity stars can provide crucial information on the chemical composition in the early Galaxy and possible internal mixing processes that can alter the surface composition of the stars. Aims. This work aims to investigate the chemical abundance patterns for CNO elements and Li in a homogeneously analyzed sample of 52 metal-poor halo giant stars. Methods. We used high-resolution spectra with a high signal-to-noise ratio (S/N) to carry out a spectral synthesis to derive detailed C, N, O, and Li abundances for a sample of stars with metallicities in the range of -3.58 <= [Fe/H] <= -1.79 dex. Our study was based on the assumption of one-dimensional (1D) local thermodynamic equilibrium (LTE) atmospheres. Results. Based on carbon and nitrogen abundances, we investigated the deep mixing taking place within stars along the red giant branch (RGB). The individual abundances of carbon decrease towards the upper RGB while nitrogen shows an increasing trend, indicating that carbon has been converted into nitrogen. No signatures of ON-cycle processed material were found for the stars in our sample. We computed a set of galactic chemical evolution (GCE) models, implementing different sets of massive star yields, both with and without including the effects of stellar rotation on nucleosynthesis. We confirm that stellar rotation is necessary to explain the highest [N/Fe] and [N/O] ratios observed in unmixed halo stars. The predicted level of N enhancement varies sensibly in dependence of the specific set of yields that are adopted. For stars with stellar parameters similar to those of our sample, heavy elements such as Sr, Y, and Zr appear to have unchanged abundances despite the stellar evolution mixing processes., Comment: 19 pages, 17 figures. Accepted for publication in A&A on Semptember 04, 2024. Table with derived abundances will be available at the CDS after the paper will be published

Run-time steering strategies like Medprompt are valuable for guiding large language models (LLMs) to top performance on challenging tasks. Medprompt demonstrates that a general LLM can be focused to deliver state-of-the-art performance on specialized domains like medicine by using a prompt to elicit a run-time strategy involving chain of thought reasoning and ensembling. OpenAI's o1-preview model represents a new paradigm, where a model is designed to do run-time reasoning before generating final responses. We seek to understand the behavior of o1-preview on a diverse set of medical challenge problem benchmarks. Following on the Medprompt study with GPT-4, we systematically evaluate the o1-preview model across various medical benchmarks. Notably, even without prompting techniques, o1-preview largely outperforms the GPT-4 series with Medprompt. We further systematically study the efficacy of classic prompt engineering strategies, as represented by Medprompt, within the new paradigm of reasoning models. We found that few-shot prompting hinders o1's performance, suggesting that in-context learning may no longer be an effective steering approach for reasoning-native models. While ensembling remains viable, it is resource-intensive and requires careful cost-performance optimization. Our cost and accuracy analysis across run-time strategies reveals a Pareto frontier, with GPT-4o representing a more affordable option and o1-preview achieving state-of-the-art performance at higher cost. Although o1-preview offers top performance, GPT-4o with steering strategies like Medprompt retains value in specific contexts. Moreover, we note that the o1-preview model has reached near-saturation on many existing medical benchmarks, underscoring the need for new, challenging benchmarks. We close with reflections on general directions for inference-time computation with LLMs., Comment: 25 pages

Games represent a significant part of modern culture, which demonstrates the importance of ensuring that everyone can participate and play in order to feel included in our society. However, most digital games end up being inaccessible to people with disabilities. Part of the problem when thinking about inclusive game design is that there is no single solution for accessibility, and what works well for one group may not work for another. This work proposes a set of guidelines for the development of inclusive mobile games, considering the widespread use of smartphones by the population and the need to include people with disabilities in the gaming culture., Comment: in Portuguese language

The Variational Quantum Eigensolver (VQE) is a promising hybrid algorithm, utilizing both quantum and classical computers to obtain the ground state energy of molecules. In this context, this study applies VQE to investigate the ground state of protocatechuic acid, analyzing its performance with various Ans\"atze and active spaces. Subsequently, all VQE results were compared to those obtained with the Hartree-Fock (HF) method. The results demonstrate that Ans\"atze, like unitary coupled-cluster singles and doubles (UCCSD) and variations of Hardware-Efficient, generally achieve accuracy close to that of HF. Furthermore, the increase in active space has led to the models becoming more difficult to converge to values with a greater distance from the correct energy. In summary, the findings of this study reinforce the use of VQE as a powerful tool for analyzing molecular ground state energies. Finally, the results underscore the critical importance of Ansatz selection and active space size in VQE performance, providing valuable insights into its potential and limitations., Comment: 10 pages, 6 figures, 1 table

The Pauli exclusion principle combined with interactions between fermions is a unifying basic mechanism that can give rise to quantum phases with spin order in diverse physical systems. Transition-metal ferromagnets, with isotropic ordering respecting crystallographic rotation symmetries and with a net magnetization, are a relatively common manifestation of this mechanism, leading to numerous practical applications, e.g., in spintronic information technologies. In contrast, superfluid $^3$He has been a unique and fragile manifestation, in which the spin-ordered phase is anisotropic, breaking the real-space rotation symmetries, and has zero net magnetization. The recently discovered altermagnets share the spin-ordered anisotropic zero-magnetization nature of superfluid $^3$He. Yet, altermagnets appear to be even more abundant than ferromagnets, can be robust, and are projected to offer superior scalability for spintronics compared to ferromagnets. Our Perspective revisits the decades of research of the spin-ordered anisotropic zero-magnetization phases including, besides superfluid $^3$He, also theoretically conceived counterparts in nematic electronic liquid-crystal phases. While all sharing the same extraordinary character of symmetry breaking, we highlight the distinctions in microscopic physics which set altermagnets apart and enable their robust and abundant material realizations., Comment: 12 pages, 4 figures

We study time-crystalline order in periodically driven quantum Ising models on disorder-free decorated lattices. Using a tensor network ansatz for the state which reflects the geometry of a unit cell of the lattice, we show through finite entanglement scaling that the system has an exponentially long-lived subharmonic response in the thermodynamic limit. The resulting discrete time crystal is not only stable to imperfections in the transverse field, but also exhibits a bipartite rigidity to generic perturbations in the longitudinal field. We call this state a bipartite discrete time crystal and reveal a rich prethermal phase diagram, including multiple regions of bipartite time-crystalline order, uniform time-crystalline order and thermalization, with boundaries depending delicately on the topology of the decorated lattice. Our results thus uncover a variety of time crystals which may be realized on current digital quantum processors and analog quantum simulators.

We consider a scalar parabolic partial differential equation on the interval with nonlinear boundary conditions that are asymptotically sublinear. As the parameter crosses critical values (e.g. the Steklov eigenvalues), it is known that there are large equilibria that arise through a bifurcation from infinity (i.e., such equilibria converge, after rescaling, to the Steklov eigenfunctions). We provide a compactification approach to the study of such unbounded bifurcation curves of equilibria, their stability, and heteroclinic orbits. In particular, we construct an induced semiflow at infinity such that the Steklov eigenfunctions are equilibria. Moreover, we prove the existence of infinite-time blow-up solutions that converge, after rescaling, to certain eigenfunctions that are equilibria of the induced semiflow at infinity., Comment: 18 pages, 8 figures

The solar magnetized corona is responsible for various manifestations with a space weather impact, such as flares, coronal mass ejections (CMEs) and, naturally, the solar wind. Modeling the corona's dynamics and evolution is therefore critical for improving our ability to predict space weather In this work, we demonstrate that generative deep learning methods, such as Denoising Diffusion Probabilistic Models (DDPM), can be successfully applied to simulate future evolutions of the corona as observed in Extreme Ultraviolet (EUV) wavelengths. Our model takes a 12-hour video of an Active Region (AR) as input and simulate the potential evolution of the AR over the subsequent 12 hours, with a time-resolution of two hours. We propose a light UNet backbone architecture adapted to our problem by adding 1D temporal convolutions after each classical 2D spatial ones, and spatio-temporal attention in the bottleneck part. The model not only produce visually realistic outputs but also captures the inherent stochasticity of the system's evolution. Notably, the simulations enable the generation of reliable confidence intervals for key predictive metrics such as the EUV peak flux and fluence of the ARs, paving the way for probabilistic and interpretable space weather forecasting. Future studies will focus on shorter forecasting horizons with increased spatial and temporal resolution, aiming at reducing the uncertainty of the simulations and providing practical applications for space weather forecasting. The code used for this study is available at the following link: https://github.com/gfrancisco20/video_diffusion

In this paper, we consider the monoids of all partial endomorphisms, of all partial weak endomorphisms, of all injective partial endomorphisms, of all partial strong endomorphisms and of all partial strong weak endomorphisms of a star graph with a finite number of vertices. Our main objective is to exhibit a presentation for each of them.

For a digraph $D$ of order $n$ and an integer $1 \leq k \leq n-1$, the $k$-token digraph of $D$ is the graph whose vertices are all $k$-subsets of vertices of $D$ and, given two such $k$-subsets $A$ and $B$, $(A,B)$ is an arc in the $k$-token digraph whenever $\{a\} = A \setminus B$, $\{b\} = B \setminus A$, and there is an arc $(a,b)$ in $D$. Token digraphs are a generalization of token graphs. In this paper, we study some properties of token digraphs, including strong and unilateral connectivity, kernels, girth, circumference and Eulerianity. We also extend some known results on the clique and chromatic numbers of $k$-token graphs, addressing the bidirected clique number and dichromatic number of $k$-token digraphs. Additionally, we prove that determining whether $2$-token digraphs have a kernel is NP-complete., Comment: 24 pages, 9 figures

Developing a reinforcement learning (RL) agent often involves identifying values for numerous parameters, covering the policy, reward function, environment, and agent-internal architecture. Since these parameters are interrelated in complex ways, optimizing them is a black-box problem that proves especially challenging for nonexperts. Although existing optimization-as-a-service platforms (e.g., Vizier and Optuna) can handle such problems, they are impractical for RL systems, since the need for manual user mapping of each parameter to distinct components makes the effort cumbersome. It also requires understanding of the optimization process, limiting the systems' application beyond the machine learning field and restricting access in areas such as cognitive science, which models human decision-making. To tackle these challenges, the paper presents AgentForge, a flexible low-code platform to optimize any parameter set across an RL system. Available at https://github.com/feferna/AgentForge, it allows an optimization problem to be defined in a few lines of code and handed to any of the interfaced optimizers. With AgentForge, the user can optimize the parameters either individually or jointly. The paper presents an evaluation of its performance for a challenging vision-based RL problem., Comment: This paper has been accepted at the 17th International Conference on Agents and Artificial Intelligence (ICAART 2025)

Solar flare forecasting mainly relies on photospheric magnetograms and associated physical features to predict forthcoming flares. However, it is believed that flare initiation mechanisms often originate in the chromosphere and the lower corona. In this study, we employ deep learning as a purely data-driven approach to compare the predictive capabilities of chromospheric and coronal UV and EUV emissions across different wavelengths with those of photospheric line-of-sight magnetograms. Our findings indicate that individual EUV wavelengths can provide discriminatory power comparable or better to that of line-of-sight magnetograms. Moreover, we identify simple multimodal neural network architectures that consistently outperform single-input models, showing complementarity between the flare precursors that can be extracted from the distinct layers of the solar atmosphere. To mitigate potential biases from known misattributions in Active Region flare catalogs, our models are trained and evaluated using full-disk images and a comprehensive flare event catalog at the full-disk level. We introduce a deep-learning architecture suited for extracting temporal features from full-disk videos.