Search

We introduce Docling, an easy-to-use, self-contained, MIT-licensed, open-source toolkit for document conversion, that can parse several types of popular document formats into a unified, richly structured representation. It is powered by state-of-the-art specialized AI models for layout analysis (DocLayNet) and table structure recognition (TableFormer), and runs efficiently on commodity hardware in a small resource budget. Docling is released as a Python package and can be used as a Python API or as a CLI tool. Docling's modular architecture and efficient document representation make it easy to implement extensions, new features, models, and customizations. Docling has been already integrated in other popular open-source frameworks (e.g., LangChain, LlamaIndex, spaCy), making it a natural fit for the processing of documents and the development of high-end applications. The open-source community has fully engaged in using, promoting, and developing for Docling, which gathered 10k stars on GitHub in less than a month and was reported as the No. 1 trending repository in GitHub worldwide in November 2024., Comment: Accepted to AAAI 25: Workshop on Open-Source AI for Mainstream Use

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 3,000 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 25 pages, 6 figures

The coherent spin dynamics of electrons and holes are studied in a FA0.9Cs0.1PbI2.8Br0.2 perovskite bulk crystal, using time-resolved Kerr ellipticity in a two-color pump-probe scheme. The probe photon energy is tuned to the exciton resonance, while the pump photon energy is detuned from it up to 0.75 eV to higher energies. The spin-oriented electrons and holes photogenerated with significant excess kinetic energy relax into states in vicinity of the band gap, where they undergo Larmor precession in an external magnetic field. At cryogenic temperatures down to 1.6 K, the spin dephasing time reaches the nanosecond range. During energy relaxation, the carrier spin relaxation is inefficient and only happens when the carriers become localized. In experiments with two pump pulses, all-optical control of the amplitudes and phases of the electron and hole spin signals is achieved in the additive regime by varying the intensities of the pump pulses and the time delay between them.

Ultrafast optical pump-probe spectroscopy is a powerful tool to study dynamics in solid materials on femto- and picosecond timescales. In such experiments, a pump pulse induces dynamics inside a sample by impulsive light-matter interaction, resulting in dynamics that can be detected using a time-delayed probe pulse. In addition to the desired dynamics, the initial interaction may also lead to unwanted effects that may result in irreversible changes and even damage. Therefore, the achievable signal strength is often limited by the pumping conditions that a sample can sustain. Here we investigate the optimization of ultrafast photoacoustics in various solid thin films. We perform systematic experiments aimed at maximizing the achievable signal-to-noise (SNR) ratio in a given measurement time while limiting sample damage. By varying pump and probe pulse energies, average pump fluence, and repetition rate, we identify different paths towards optimal SNR depending on material properties. Our results provide a strategy for the design of pump-probe experiments, to optimize achievable SNR for samples in which different damage mechanisms may dominate., Comment: 9 pages, 10 figures

We present a new, fully consistent framework for describing GeV-scale axion-like particles (ALPs) coupled to gluons. Unlike previous approaches, which either rely on pure chiral perturbation theory or introduce scheme-dependent parameters, our treatment simultaneously (i) enforces invariance under unphysical chiral rotations used to remove the ALP-gluon coupling, and (ii) incorporates heavier resonances relevant at the GeV scale. This construction enables robust matching between exclusive and perturbative descriptions of ALP decays, and reveals that commonly used descriptions can underestimate or overestimate ALP decay widths and production rates by orders of magnitude. We further show that the production of ALPs at proton accelerator experiments can incur large theoretical uncertainties, significantly impacting ALP event yields at past and future Lifetime Frontier experiments., Comment: 5+5 pages

We study topologization of the semigroup $\mathscr{O\!\!I}\!_n(L)$ of finite partial order isomorphisms of a bounded rank of an infinite linear ordered set $(L,\leqslant)$. In particular we show that every $T_1$ left-topological (right-topological) semigroup $\mathscr{O\!\!I}\!_n(L)$ is a completely Hausdorff, Urysohn, totally separated, scattered space. We prove that on the semigroup $\mathscr{O\!\!I}\!_n(L)$ admits a unique Hausdorff countably compact (pseudocompact) shift-continuous topology which is compact, and the Bohr compactification of a Hausdorff topological semigroup $\mathscr{O\!\!I}\!_n(L)$ is the trivial semigroup., Comment: 9 pages (in Ukrainian)

We find all K-polystable limits of divisors in $(\mathbb{P}^1)^4$ of degree $(1,1,1,1)$ and explicitly describe the associated irreducible component of the K-moduli space., Comment: 28 pages

The Ni-Au-Ga interdiffusion mechanisms taking place during rapid thermal annealing (RTA) under oxygen atmosphere of a Ni-Au/p-GaN contact are investigated by high-resolution transmission electron microscopy (HR-TEM) coupled to energy dispersive X-ray spectroscopy (EDX). It is shown that oxygen-assisted, Ni diffusion to the top surface of the metallic contact through the formation of a nickel oxide (NiOx) is accompanied by Au diffusion down to the GaN surface, and by Ga out-diffusion through the GaN/metal interface. Electrical characterizations of the contact by Transmission Line Method (TLM) show that an ohmic contact is obtained as soon as a thin, Au-Ga interfacial layer is formed, even after complete diffusion of Ni or NiOx to the top surface of the contact. Our results clarify that the presence of Ni or NiOx at the interface is not the main origin of the ohmic-like behavior in such contacts. Auto-cleaning of the interface during the interdiffusion process may play a role, but TEM-EDX analysis evidences that the creation of Ga vacancies associated to the formation of a Ga-Au interfacial layer is crucial for reducing the Schottky barrier height, and maximizing the amount of current flowing through the contact.

This paper describes applying manifold learning, the novel technique of dimensionality reduction, to the images of the Galaxy Zoo DECaLs database with the purpose of building an unsupervised learning model for galaxy morphological classification. The manifold learning method assumes that data points can be projected from a manifold in high-dimensional space to a lower-dimensional Euclidean one while maintaining proximity between the points. In our case, data points are photos of galaxies from the Galaxy Zoo DECaLs database, which consists of more than 300,000 human-labeled galaxies of different morphological types. The dimensionality of such data points is equal to the number of pixels in a photo, so dimensionality reduction becomes a handy idea to help one with the successive clusterization of the data. We perform it using Locally Linear Embedding, a manifold learning algorithm, designed to deal with complex high-dimensional manifolds where the data points are originally located. After the dimensionality reduction, we perform the classification procedure on the dataset. In particular, we train our model to distinguish between round and cigar-shaped elliptical galaxies, smooth and featured spiral galaxies, and galaxies with and without disks viewed edge-on. In each of these cases, the number of classes is pre-determined. The last step in our pipeline is k-means clustering by silhouette or elbow method in lower-dimensional space. In the final case of unsupervised classification of the whole dataset, we determine that the optimal number of morphological classes of galaxies coincides with the number of classes defined by human astronomers, further confirming the feasibility and efficiency of manifold learning for this task., Comment: 7 pages, 11 figures, submitted to Astronomy and Computing Journal

Copyright infringement in frontier LLMs has received much attention recently due to the New York Times v. OpenAI lawsuit, filed in December 2023. The New York Times claims that GPT-4 has infringed its copyrights by reproducing articles for use in LLM training and by memorizing the inputs, thereby publicly displaying them in LLM outputs. Our work aims to measure the propensity of OpenAI's LLMs to exhibit verbatim memorization in its outputs relative to other LLMs, specifically focusing on news articles. We discover that both GPT and Claude models use refusal training and output filters to prevent verbatim output of the memorized articles. We apply a basic prompt template to bypass the refusal training and show that OpenAI models are currently less prone to memorization elicitation than models from Meta, Mistral, and Anthropic. We find that as models increase in size, especially beyond 100 billion parameters, they demonstrate significantly greater capacity for memorization. Our findings have practical implications for training: more attention must be placed on preventing verbatim memorization in very large models. Our findings also have legal significance: in assessing the relative memorization capacity of OpenAI's LLMs, we probe the strength of The New York Times's copyright infringement claims and OpenAI's legal defenses, while underscoring issues at the intersection of generative AI, law, and policy.

Large Language Models (LLMs) have shown impressive results in multiple domains of natural language processing (NLP) but are mainly focused on the English language. Recently, more LLMs have incorporated a larger proportion of multilingual text to represent low-resource languages. In Arabic NLP, several Arabic-centric LLMs have shown remarkable results on multiple benchmarks in the past two years. However, most Arabic LLMs have more than 7 billion parameters, which increases their hardware requirements and inference latency, when compared to smaller LLMs. This paper introduces Arabic Stable LM 1.6B in a base and chat version as a small but powerful Arabic-centric LLM. Our Arabic Stable LM 1.6B chat model achieves impressive results on several benchmarks beating multiple models with up to 8x the parameters. In addition, we show the benefit of mixing in synthetic instruction tuning data by augmenting our fine-tuning data with a large synthetic dialogue dataset.

Technology involving hybrid superconductor-semiconductor materials is a promising avenue for engineering quantum devices for information storage, manipulation, and transmission. Proximity-induced superconducting correlations are an essential part of such devices. While the proximity effect in the conduction band of common semiconductors is well understood, its manifestation in confined hole gases, realized for instance in germanium, is an active area of research. Lower-dimensional hole-based systems, particularly in germanium, are emerging as an attractive platform for a variety of solid-state quantum devices, due to their combination of efficient spin and charge control and long coherence times. The recent experimental realization of the proximity effect in germanium thus calls for a theoretical description that is tailored to hole gases. In this work, we propose a simple model to describe proximity-induced superconductivity in two-dimensional hole gases, incorporating both the heavy-hole (HH) and light-hole (LH) bands. We start from the Luttinger-Kohn model, introduce three parameters that characterize hopping across the superconductor-semiconductor interface, and derive explicit intraband and interband effective pairing terms for the HH and LH bands. Unlike previous approaches, our theory provides a quantitative relationship between induced pairings and interface properties. Restricting our general model to an experimentally relevant case where only the HH band crosses the chemical potential, we predict the coexistence of $s$-wave and $d$-wave singlet pairings, along with triplet-type pairings, and modified Zeeman and Rashba spin-orbit couplings. Our model thus presents a starting point for theoretical modeling of quantum devices based on proximitized hole gases, fueling further progress in quantum technology., Comment: 22 pages, 5 figures

Efficiency and emission rate are two traditionally conflicting parameters in radiation detection, and achieving their simultaneous maximization could significantly advance ultrafast time-of-flight (ToF) technologies. In this study, we demonstrate that this goal is attainable by harnessing the giant oscillator strength (GOS) inherent to weakly confined perovskite nanocrystals, which enables superradiant scintillation under mildly cryogenic conditions that align seamlessly with ToF technologies. We show that the radiative acceleration due to GOS encompasses both single and multiple exciton dynamics arising from ionizing interactions, further enhanced by suppressed non-radiative losses and Auger recombination at 80 K. The outcome is ultrafast scintillation with 420 ps lifetime and light yield of ~10'000 photons/MeV for diluted NC solutions, all without non-radiative losses. Temperature-dependent light-guiding experiments on test-bed nanocomposite scintillators finally indicate that the light-transport capability remains unaffected by the accumulation of band-edge oscillator strength due to GOS. These findings suggest a promising pathway toward developing ultrafast nanotech scintillators with optimized light output and timing performance.

Economic growth in Sweden during the early 20th Century was largely driven by industry. A significant contributor to this growth was the installation of different kinds of engines used to power factories. We use newly digitized data on engines and their energy source by industry sector, and combine this with municipality-level data of workers per industry sector to construct a new variable reflecting economic output using dirty engines. In turn, we assess the average externality of dirty output on mortality in the short-run, as defined by deaths over the population in the baseline year. Our results show substantial increases of up to 17% higher mortality in cities where large increases to dirty engine installations occurred, which is largely driven by the elderly. We also run a placebo test using clean powered industry and find no effect on mortality., Comment: 23 pages, 15 figures, 9 tables

Frontier AI safety policies highlight automation of AI research and development (R&D) by AI agents as an important capability to anticipate. However, there exist few evaluations for AI R&D capabilities, and none that are highly realistic and have a direct comparison to human performance. We introduce RE-Bench (Research Engineering Benchmark, v1), which consists of 7 challenging, open-ended ML research engineering environments and data from 71 8-hour attempts by 61 distinct human experts. We confirm that our experts make progress in the environments given 8 hours, with 82% of expert attempts achieving a non-zero score and 24% matching or exceeding our strong reference solutions. We compare humans to several public frontier models through best-of-k with varying time budgets and agent designs, and find that the best AI agents achieve a score 4x higher than human experts when both are given a total time budget of 2 hours per environment. However, humans currently display better returns to increasing time budgets, narrowly exceeding the top AI agent scores given an 8-hour budget, and achieving 2x the score of the top AI agent when both are given 32 total hours (across different attempts). Qualitatively, we find that modern AI agents possess significant expertise in many ML topics -- e.g. an agent wrote a faster custom Triton kernel than any of our human experts' -- and can generate and test solutions over ten times faster than humans, at much lower cost. We open-source the evaluation environments, human expert data, analysis code and agent trajectories to facilitate future research.

Computer vision tasks require processing large amounts of data to perform image classification, segmentation, and feature extraction. Optical preprocessors can potentially reduce the number of floating point operations required by computer vision tasks, enabling low-power and low-latency operation. However, existing optical preprocessors are mostly learned and hence strongly depend on the training data, and thus lack universal applicability. In this paper, we present a metaoptic imager, which implements the Radon transform obviating the need for training the optics. High quality image reconstruction with a large compression ratio of 0.6% is presented through the use of the Simultaneous Algebraic Reconstruction Technique. Image classification with 90% accuracy is presented on an experimentally measured Radon dataset through neural network trained on digitally transformed images.

We study higher uniformity properties of the von Mangoldt function $\Lambda$, the M\"obius function $\mu$, and the divisor functions $d_k$ on short intervals $(x,x+H]$ for almost all $x \in [X, 2X]$. Let $\Lambda^\sharp$ and $d_k^\sharp$ be suitable approximants of $\Lambda$ and $d_k$, $G/\Gamma$ a filtered nilmanifold, and $F\colon G/\Gamma \to \mathbb{C}$ a Lipschitz function. Then our results imply for instance that when $X^{1/3+\varepsilon} \leq H \leq X$ we have, for almost all $x \in [X, 2X]$, \[ \sup_{g \in \text{Poly}(\mathbb{Z} \to G)} \left| \sum_{x < n \leq x+H} (\Lambda(n)-\Lambda^\sharp(n)) \overline{F}(g(n)\Gamma) \right| \ll H\log^{-A} X \] for any fixed $A>0$, and that when $X^{\varepsilon} \leq H \leq X$ we have, for almost all $x \in [X, 2X]$, \[ \sup_{g \in \text{Poly}(\mathbb{Z} \to G)} \left| \sum_{x < n \leq x+H} (d_k(n)-d_k^\sharp(n)) \overline{F}(g(n)\Gamma) \right| = o(H \log^{k-1} X). \] As a consequence, we show that the short interval Gowers norms $\|\Lambda-\Lambda^\sharp\|_{U^s(X,X+H]}$ and $\|d_k-d_k^\sharp\|_{U^s(X,X+H]}$ are also asymptotically small for any fixed $s$ in the same ranges of $H$. This in turn allows us to establish the Hardy-Littlewood conjecture and the divisor correlation conjecture with a short average over one variable. Our main new ingredients are type $II$ estimates obtained by developing a "contagion lemma" for nilsequences and then using this to "scale up" an approximate functional equation for the nilsequence to a larger scale. This extends an approach developed by Walsh for Fourier uniformity., Comment: 104 pages

The great variety of lead halide perovskite semiconductors represents an outstanding platform for studying crystal symmetry effects on the spin-dependent properties. Access to them is granted through optical orientation of exciton and carrier spins by circularly polarized photons. Here, the exciton spin polarization is investigated at $1.6$\,K cryogenic temperature in four lead halide perovskite crystals with different symmetries: (almost) cubic in FA$_{0.9}$Cs$_{0.1}$PbI$_{2.8}$Br$_{0.2}$ and FAPbBr$_3$, and orthorhombic in MAPbI$_3$ and CsPbBr$_3$. Giant optical orientation of 85\% is found for the excitons in FA$_{0.9}$Cs$_{0.1}$PbI$_{2.8}$Br$_{0.2}$, MAPbI$_3$, and CsPbBr$_3$, while it amounts to 20\% in FAPbBr$_3$. For all studied crystals, the optical orientation is robust to detuning of the laser photon energy from the exciton resonance, remaining constant for high energy detunings up to 0.3~eV, above which it continuously decreases to zero for detunings exceeding 1~eV. No acceleration of the spin relaxation for excitons with large kinetic energy is found in the cubic and orthorhombic crystals. This evidences the absence of the Dyakonov-Perel spin relaxation mechanism, which is based on the Rashba-Dresselhaus splitting of spin states at finite $k$-vectors. This indicates that the spatial inversion symmetry is maintained in perovskite crystals, independent of the cubic or orthorhombic phase.

We study algebraic properties of the semigroup $\mathscr{O\!\!I\!}_n(L)$ of finite partial order isomorphisms of the rank $\leq n$ of an infinite linearly ordered set $(L,\leqslant)$. In particular we describe its idempotents, the natural partial order and Green's relations on $\mathscr{O\!\!I\!}_n(L)$. It is proved that the semigroup $\mathscr{O\!\!I\!}_n(L)$ is stable and it contains tight ideal series. Moreover, we show that the semigroup $\mathscr{O\!\!I\!}_n(L)$ admits only Rees' congruences and every its homomorphic image is a semigroup with tight ideal series., Comment: 7 pages (in Ukrainian)

e investigate the cosmological impact of hypothetical unstable new physics particles that decay in the MeV-scale plasma of the Early Universe. Focusing on scenarios where the decays produce metastable species such as muons, pions, and kaons, we systematically analyze the dynamics of these particles using coupled Boltzmann equations governing their abundances. Our results demonstrate that the metastable species can efficiently annihilate or interact with nucleons, which often leads to their disappearance prior to decay. The suppression of decay significantly alters the properties of cosmic neutrinos, impacting cosmological observables like Big Bang nucleosynthesis and the Cosmic Microwave Background. To support further studies, we provide a public Mathematica code that traces the evolution of these metastable particles and apply it to several new physics models., Comment: Companion paper of "New physics decaying into metastable particles: impact on cosmic neutrinos"

We investigate decays of hypothetical unstable new physics particles into metastable species such as muons, pions, or kaons in the early Universe, when temperatures are in the MeV range, and study how they affect cosmic neutrinos. We demonstrate that decays of the metastable particles compete with their annihilations and interactions with nucleons, which reduces the production of high-energy neutrinos and increases energy injection into the electromagnetic sector. This energy reallocation alters the impact of the new physics particles on the effective number of neutrino degrees of freedom, $N_{\text{eff}}$, modifies neutrino spectral distortions, and may induce asymmetries in neutrino and antineutrino energy distributions. These modifications have important implications for observables such as Big Bang Nucleosynthesis and the Cosmic Microwave Background, especially in light of upcoming CMB observations aiming to reach percent-level precision on $N_{\rm eff}$. We illustrate our findings with a few examples of new physics particles and provide a computational tool available for further exploration., Comment: Companion paper of "Dynamics of metastable Standard Model particles from long-lived particle decays in the MeV primordial plasma"

We present a detailed imaging analysis of 260 ks of sub-arcsecond resolution Chandra Advanced CCD Imaging Spectrometer (ACIS-S) observations of the nearby Seyfert 2 galaxy NGC 5728. Our study focuses on the bright and diffuse soft X-ray emission within the galaxy's inner ~1 kpc. By comparing the X-ray emission across different energy bands, we identify localized variations in the absorbing column and emission processes. We observe more X-ray absorption in the direction perpendicular to the bicone, which is co-located with an inner warped CO disk in the galaxy. The innermost region, which shows the strongest excess of hard X-ray emission, is spatially coincident with the CO(2-1) emission from ALMA and dusty spirals observed in a Hubble Space Telescope V-H color map. We detect soft extended emission associated with the circumnuclear star-forming ring at ~1 kpc, suggestive of hot gas with kT=0.44 keV. We derive measurements for the hot gas mass, M=7.9x10^5 solar masses, pressure, p=2.0x10^-10 dyne per square cm, and cooling times, t=193.2 Myr. In the vicinity of the star-forming ring, we detect two X-ray point sources with soft X-ray spectra and 0.3-7 keV luminosities L~8x10^38 erg per second. These properties suggest X-ray binaries., Comment: Accepted for publication on ApJ

We study a flaring activity of the HSP Mrk421 that was characterized from radio to very-high-energy (VHE; E $>0.1$TeV) gamma rays with MAGIC, Fermi-LAT, Swift, XMM-Newton and several optical and radio telescopes. These observations included, for the first time for a gamma-ray flare of a blazar, simultaneous X-ray polarization measurements with IXPE. We find substantial variability in both X-rays and VHE gamma rays throughout the campaign, with the highest VHE flux above 0.2 TeV occurring during the IXPE observing window, and exceeding twice the flux of the Crab Nebula. However, the VHE and X-ray spectra are on average softer, and the correlation between these two bands weaker that those reported in previous flares of Mrk421. IXPE reveals an X-ray polarization degree significantly higher than that at radio and optical frequencies. The X-ray polarization angle varies by $\sim$100$^\circ$ on timescales of days, and the polarization degree changes by more than a factor 4. The highest X-ray polarization degree reaches 26%, around which a X-ray counter-clockwise hysteresis loop is measured with XMM-Newton. It suggests that the X-ray emission comes from particles close to the high-energy cutoff, hence possibly probing an extreme case of the Turbulent Extreme Multi-Zone model. We model the broadband emission with a simplified stratified jet model throughout the flare. The polarization measurements imply an electron distribution in the X-ray emitting region with a very high minimum Lorentz factor, which is expected in electron-ion plasma, as well as a variation of the emitting region size up to a factor of three during the flaring activity. We find no correlation between the fluxes and the evolution of the model parameters, which indicates a stochastic nature of the underlying physical mechanism. Such behaviour would be expected in a highly turbulent electron-ion plasma crossing a shock front., Comment: Accepted in Astronomy and Astrophysics (on January 17th 2025). Corresponding authors: Axel Arbet-Engels, Lea Heckmann, David Paneque

Recently published work on rephrasing natural text data for pre-training LLMs has shown promising results when combining the original dataset with the synthetically rephrased data. We build upon previous work by replicating existing results on C4 and extending them with our optimized rephrasing pipeline to the English, German, Italian, and Spanish Oscar subsets of CulturaX. Our pipeline leads to increased performance on standard evaluation benchmarks in both the mono- and multilingual setup. In addition, we provide a detailed study of our pipeline, investigating the choice of the base dataset and LLM for the rephrasing, as well as the relationship between the model size and the performance after pre-training. By exploring data with different perceived quality levels, we show that gains decrease with higher quality. Furthermore, we find the difference in performance between model families to be bigger than between different model sizes. This highlights the necessity for detailed tests before choosing an LLM to rephrase large amounts of data. Moreover, we investigate the effect of pre-training with synthetic data on supervised fine-tuning. Here, we find increasing but inconclusive results that highly depend on the used benchmark. These results (again) highlight the need for better benchmarking setups. In summary, we show that rephrasing multilingual and low-quality data is a very promising direction to extend LLM pre-training data., Comment: 21 pages, 4 figures, 12 tables

X-ray polarization is a unique new probe of the particle acceleration in astrophysical jets made possible through the Imaging X-ray Polarimetry Explorer. Here we report on the first dense X-ray polarization monitoring campaign on the blazar Mrk 421. Our observations were accompanied by an even denser radio and optical polarization campaign. We find significant short-timescale variability in both X-ray polarization degree and angle, including a $\sim90^\circ$ angle rotation about the jet axis. We attribute this to random variations of the magnetic field, consistent with the presence of turbulence but also unlikely to be explained by turbulence alone. At the same time, the degree of lower-energy polarization is significantly lower and shows no more than mild variability. Our campaign provides further evidence for a scenario in which energy-stratified shock-acceleration of relativistic electrons, combined with a turbulent magnetic field, is responsible for optical to X-ray synchrotron emission in blazar jets., Comment: 23 pages, including 8 pages of appendices. 12 figures, 3 tables. Submitted to ApJ

Persistent revivals recently observed in Rydberg atom simulators have challenged our understanding of thermalization and attracted much interest to the concept of quantum many-body scars (QMBSs). QMBSs are non-thermal highly excited eigenstates that coexist with typical eigenstates in the spectrum of many-body Hamiltonians, and have since been reported in multiple theoretical models, including the so-called PXP model, approximately realized by Rydberg simulators. At the same time, questions of how common QMBSs are and in what models they are physically realized remain open. In this Letter, we demonstrate that QMBSs exist in a broader family of models that includes and generalizes PXP to longer-range constraints and states with different periodicity. We show that in each model, multiple QMBS families can be found. Each of them relies on a different approximate $\mathfrak{su}(2)$ algebra, leading to oscillatory dynamics in all cases. However, in contrast to the PXP model, their observation requires launching dynamics from weakly entangled initial states rather than from a product state. QMBSs reported here may be experimentally probed using Rydberg atom simulator in the regime of longer-range Rydberg blockades., Comment: 8+12 pages, 6+8 figures

We use the Hubble Space Telescope to analyse the extended [O III] 5007A emission in seven bright radio-quiet type 1 quasars (QSO1s), focusing on the morphology and physical conditions of their extended Narrow-Line Regions (NLRs). We find NLRs extending 3-9 kpc, with four quasars showing roughly symmetrical structures (b/a=1.2-1.5) and three displaying asymmetric NLRs (b/a=2.4-5.6). When included with type 1 and type 2 AGNs from previous studies, the sizes of the extended [O III] regions scale with luminosity as $R[O III] \sim L[O III]^{0.5}$, consistent with photoionisation. However, when analysed separately, type 1s exhibit a steeper slope ($\gamma=0.57\pm0.05$) compared to type 2 AGNs ($\gamma=0.48\pm0.02$). We use photoionisation modeling to estimate the maximum NLRs sizes, assuming a minimum ionisation parameter of $\log(U) = -3$, an ionising luminosity based on the $L[O III]$-derived bolometric luminosity, and a minimum gas number density $n_H \sim 100\,\text{cm}^{-3}$, assuming that molecular clouds provide a reservoir for the ionised gas. The derived sizes agree well with direct measurements for a sample of type 2 quasars, but are underestimated for the current sample of QSO1s. A better agreement is obtained for the QSO1s using bolometric luminosities derived from the 5100A continuum luminosity. Radial mass profiles for the QSO1s show significant extended mass in all cases, but with less [O III]-emitting gas near the central AGN compared to QSO2s. This may suggest that the QSO1s are in a later evolutionary stage than QSO2s, further past the blow-out stage., Comment: Accepted for publication on MNRAS

The Standardized Uptake Value (SUV) is a critical metric in positron emission tomography (PET) imaging, used to assess metabolic activity. However, calculating SUV from DICOM files presents challenges due to vendor-specific DICOM attributes and variations in the encoding of radiotracer accumulation times. This technical note introduces a robust, vendor-specific SUV calculation strategy that addresses inconsistencies in current methodologies. We also integrate this strategy into an open-source software solution, Z-Rad, capable of converting raw PET DICOM data into body-weight normalized SUV NIfTI files. Our SUV calculation strategy was developed by reviewing DICOM conformance statements from GE, Philips, and Siemens. Validation was conducted using real-world PET datasets, and the proposed strategy was compared to existing software solutions. Our SUV calculation approach demonstrated improved accuracy, particularly in resolving time-related discrepancies in the studied data. Our analysis also identified inconsistencies in the SUV calculation methods used by popular commercial and open-source software solutions, which do not fully account for vendor-specific DICOM attributes and PET image acquisition times. These limitations resulted in errors in SUV estimation reaching 33\% when comparing our strategy to studied software. The proposed vendor-specific SUV calculation strategy significantly enhances accuracy in PET imaging by addressing key inconsistencies caused by variations in DICOM attributes and image acquisition times across different vendors. This method effectively reduces SUV calculation errors and has been integrated into an open-source software, Z-Rad.

The robustness of LLMs to jailbreak attacks, where users design prompts to circumvent safety measures and misuse model capabilities, has been studied primarily for LLMs acting as simple chatbots. Meanwhile, LLM agents -- which use external tools and can execute multi-stage tasks -- may pose a greater risk if misused, but their robustness remains underexplored. To facilitate research on LLM agent misuse, we propose a new benchmark called AgentHarm. The benchmark includes a diverse set of 110 explicitly malicious agent tasks (440 with augmentations), covering 11 harm categories including fraud, cybercrime, and harassment. In addition to measuring whether models refuse harmful agentic requests, scoring well on AgentHarm requires jailbroken agents to maintain their capabilities following an attack to complete a multi-step task. We evaluate a range of leading LLMs, and find (1) leading LLMs are surprisingly compliant with malicious agent requests without jailbreaking, (2) simple universal jailbreak templates can be adapted to effectively jailbreak agents, and (3) these jailbreaks enable coherent and malicious multi-step agent behavior and retain model capabilities. To enable simple and reliable evaluation of attacks and defenses for LLM-based agents, we publicly release AgentHarm at https://huggingface.co/datasets/ai-safety-institute/AgentHarm.

Motivated by the back-reaction debate, and~some unexplained characteristics of the CMB, we investigate the possibility of some anisotropy in the universe observed around us. To this aim, we build up a novel prediction for the Hubble law for the late universe from a Bianchi type I model, {taken as proof of concept,} transcribing the departure of such model from a $\Lambda$CDM model. We~dicussed the redshift measurement in this universe, and~finally formalized the Hubble diagram., Comment: 17pp 1fig, published in Symmetry

Neutron stars change their structure with accumulation of dark matter. We study how their mass is influenced from the environment. Close to the sun, the dark matter accretion from the neutron star does not have any effect on it. Moving towards the galactic center, the density increase in dark matter results in increased accretion. At distances of some fraction of a parsec, the neutron star acquire enough dark matter to have its structure changed. We show that the neutron star mass decreases going towards the galactic centre, and that dark matter accumulation beyond a critical value collapses the neutron star into a black hole. Calculations cover cases varying the dark matter particle mass, self-interaction strength, and ratio between the pressure of dark matter and ordinary matter. This allow us to constrain the interaction cross section, $\sigma_{\rm dm}$, between nucleons and dark matter particles, as well as the dark matter self-interaction cross section., Comment: 15pp 5figs, published in Universe

Interactions between emitters within an ensemble can give rise to cooperative processes that significantly alter the properties of the emitted light. One such process is superfluorescence (SF), where excited electric dipoles spontaneously couple coherently and effectively radiate as one macroscopic emitter. It requires low energetic disorder, high temporal coherence and oscillator strength, and sub-wavelength volumes of material can be sufficient. Conversely, amplified spontaneous emission (ASE) originates from an avalanche-like stimulated amplification of initially spontaneously emitted photons and does not necessitate temporally coherent interactions among the emitters, but rather requires spatially long enough light propagation within the material to harvest the optical gain. Cesium lead halide perovskite nanocrystals (NCs) are one of the very few materials where both ASE (in disordered films) and SF (in ordered assemblies) were observed, however leaving unclear whether and how these regimes could be connected. Here, we demonstrate that temperature and excitation density can drive the transition between both regimes in a thin film of giant CsPbBr3 perovskite NCs. At temperatures below 45 K, excitonic SF was observed, whereas above a transition range between 45 K and 100 K, ASE prevails, but requires increased optical excitation and emitter density. Our results work out the different collective effects present in lead halide perovskites, providing fundamental insights into cooperative phenomena and important guidance for the development of compact and bright (quantum) light sources.

Machine learning has emerged as a potent computational tool for expediting research and development in solid oxide fuel cell electrodes. The effective application of machine learning for performance prediction requires transforming electrode microstructure into a format compatible with artificial neural networks. Input data may range from a comprehensive digital material representation of the electrode to a selected set of microstructural parameters. The chosen representation significantly influences the performance and results of the network. Here, we show a novel approach utilizing persistence representation derived from computational topology. Using 500 microstructures and current-voltage characteristics obtained with 3D first-principles simulations, we have prepared an artificial neural network model that can replicate current-voltage characteristics of unseen microstructures based on their persistent image representation. The artificial neural network can accurately predict the polarization curve of solid oxide fuel cell electrodes. The presented method incorporates complex microstructural information from the digital material representation while requiring substantially less computational resources (preprocessing and prediction time approximately 1 min) compared to our high-fidelity simulations (simulation time approximately 1 hour) to obtain a single current-potential characteristic for one microstructure., Comment: 20 pages, 12 figures, submitted to Energy and AI

Blazars, a class of active galactic nuclei (AGN) powered by supermassive black holes, are known for their remarkable variability across multiple timescales and wavelengths. With advancements in both ground- and space-based telescopes, our understanding of AGN central engines has significantly improved. However, the mechanisms driving this variability remain elusive, and continue to fascinate both theorists and observers alike. The primary objective of this study is to constrain the X-ray variability properties of the TeV blazar PKS 2155-304. We conduct a comprehensive X-ray spectral and timing analysis, focusing on both long-term and intra-day variability. This analysis uses data from 22 epochs of XMM-Newton EPIC-pn observations, collected over 15 years (2000-2014). To investigate the variability of the source, we applied both timing and spectral analyses. For the timing analysis, we estimated fractional variability, variability amplitude, minimum variability timescales, flux distribution, and power spectral density (PSD). In the spectral analysis, we fitted the X-ray spectra using power-law, log-parabola, and broken power-law (BPL) models to determine the best-fitting parameters. Additionally, we studied the hardness ratio (HR). We observed moderate intra-day variability in most of the light curves. Seven out of the twenty-two observations showed a clear bimodal flux distribution, indicating the presence of two distinct flux states. Our analysis revealed a variable power-law PSD slope. Most HR plots did not show significant variation with flux, except for one observation (OBSID 0124930501), where HR increased with flux (Count/s). The fitted X-ray spectra favored the BPL model for the majority of observations. The findings of this work shed light on the intraday variability of blazars, providing insights into the non-thermal jet processes that drive the observed flux variations.

The recent emergence of quantum confined nanomaterials in the field of radiation detection, in particular lead halide perovskite nanocrystals, offers potentially revolutionary scalability and performance advantages over conventional materials. This development raises fundamental questions about the mechanism of scintillation itself at the nanoscale and the role of particle size, arguably the most defining parameter of quantum dots. Understanding this is crucial for the design and optimisation of future nanotechnology scintillators. In this work, we address these open questions by theoretically and experimentally studying the size-dependent scintillation of CsPbBr3 nanocrystals using a combination of Monte Carlo simulations, spectroscopic, and radiometric techniques. The results reveal and unravel a complex parametric space where the fine balance between the simultaneous effects of size-dependent energy deposition, (multi-)exciton population, and light emission under ionizing excitation, typical of confined particles, combine to maximize the scintillation efficiency and time performance of larger nanocrystals due to greater stopping power and reduced Auger decay. The remarkable agreement between theory and experiment produces a fully validated descriptive model that unprecedentedly predicts the scintillation yield and kinetics of nanocrystals without free parameters, providing the first fundamental guide for the rational design of nanoscale scintillators.

We show that if $\Gamma$ is a co-compact arithmetic lattice in $SL(2,\mathbb{R})$ or $\Gamma=SL(2,\mathbb{Z})$ then the horocycle orbit of every non-periodic point $x\in SL(2,\mathbb{R})/\Gamma$ equidistributes (with respect to Haar measure) when sampled at integers having exactly two prime factors.

Understanding how hypothetical new physics influences primordial neutrinos is essential in light of interpreting past and ongoing CMB observations. To reach this goal, we present a novel approach to solving the neutrino Boltzmann equation in the Early Universe. It overcomes the limitations of existing methods by providing a model-independent, computationally efficient, and accurate framework capable of handling complex interactions and non-equilibrium neutrino distributions. We demonstrate its comprehensive applicability through several case studies. In particular, we resolve the existing discrepancy between state-of-the-art approaches regarding the dynamics of the number of ultrarelativistic degrees of freedom in the presence of injections of high-energy neutrinos.

The use of a reconfigurable intelligent surface (RIS) for aiding user-specific transmission has been widely explored. However, little attention has been devoted to utilizing RIS for assisting cell-specific transmission, where the RIS needs to reflect signals in a broad angular range. Furthermore, although modern communication systems operate in two polarizations, the majority of the works on RIS consider a uni-polarized surface, only reflecting the signals in one polarization. To fill these gaps, we study a downlink broadcasting scenario where a base station (BS) sends a cell-specific signal to all the users residing at unknown locations with the assistance of a dual-polarized RIS. We utilize the duality between the auto-correlation function and power spectrum in the space/spatial frequency domain to design configurations for broad-beam reflection. We first consider a free-space line-of-sight BS-RIS channel and show that the RIS configuration matrices must form a Golay complementary array pair for broad-beam radiation. We also present how to form Golay complementary array pairs based on known Golay complementary sequence pairs. We then consider an arbitrary BS-RIS channel and propose an algorithm based on stochastic optimization to find RIS configurations that produce a practically broad beam by relaxing the requirement on uniform broadness. Numerical simulations are finally conducted to corroborate the analyses and evaluate the performance., Comment: Accepted for publication in IEEE Transactions on Wireless Communications

A systematic and unified analysis of the capability of lifetime frontier experiments to probe the parameter space of hypothetical long-lived particles is crucial for defining the targets of future searches. However, such a comprehensive study has not yet been conducted for dark photons - hypothetical massive particles that kinetically mix with Standard Model photons. Existing studies often rely on outdated assumptions about dark photon phenomenology, leading to inaccurate calculations of the parameter space, including regions already excluded or relevant for future experiments. In this paper, we present a comprehensive calculation of the parameter space for GeV-scale dark photons and light dark matter accessible to lifetime frontier experiments. Our analysis highlights how theoretical uncertainties in dark photon phenomenology can significantly impact experimental sensitivity across coupling and mass parameters. To facilitate further investigation, we provide these results in a publicly accessible form, incorporating updated dark photon phenomenology into the event generator \texttt{SensCalc}., Comment: version accepted for the publication in PRD

Cosmological observations from Big Bang Nucleosynthesis and the Cosmic Microwave Background (CMB) offer crucial insights into the Early Universe, enabling us to trace its evolution back to lifetimes as short as 0.01 seconds. Upcoming CMB spectrum measurements, such as those underway at the Simons Observatory, will achieve unprecedented precision, allowing for more accurate extraction of information about the properties of the primordial plasma and, in particular, primordial neutrinos. This provides an opportunity to test whether these properties align with the predictions of the standard cosmological model or indicate the presence of new physics that influenced the evolution of the MeV-temperature plasma. A key component in understanding how new physics may have affected primordial neutrinos is solving the neutrino Boltzmann equation. In this paper, we present a novel approach to solving this equation that offers model independence, transparency, and computational efficiency - features that current state-of-the-art methods lack. We demonstrate a proof-of-concept implementation and apply it to several toy scenarios, showcasing key aspects of the primordial plasma's evolution in the presence of new physics., Comment: Text polishing, references adding, conclusions unchanced

The discovery of an anomalous Hall effect (AHE) sensitive to the magnetic state of antiferromagnets can trigger a new era of spintronics, if materials that host a tunable and strong AHE are identified. Altermagnets are a new class of materials that can under certain conditions manifest a strong AHE, without having a net magnetization. But the ability to control their AHE is still lacking. In this study, we demonstrate that the AHE in altermagnetic {\alpha}-MnTe grown on GaAs(111) substrates can be "written on-demand" by cooling the material under an in-plane magnetic field. The magnetic field controls the strength and the coercivity of the AHE. Remarkably, this control is unique to {\alpha}-MnTe grown on GaAs and is absent in {\alpha}-MnTe grown on SrF2. The tunability that we reveal challenges our current understanding of the symmetry-allowed AHE in this material and opens new possibilities for the design of altermagnetic spintronic devices.

In 1970, Huxley obtained a sharp upper bound for the sixth moment of Dirichlet $L$-functions at the central point, averaged over primitive characters $\chi$ modulo $q$ and all moduli $q \leq Q$. In 2007, as an application of their ``asymptotic large sieve'', Conrey, Iwaniec and Soundararajan showed that when an additional short $t$-averaging is introduced into the problem, an asymptotic can be obtained. In this paper we show that this extraneous averaging can be removed, and we thus obtain an asymptotic for the original moment problem considered by Huxley. The main new difficulty in our work is the appearance of certain challenging ``unbalanced'' sums that arise as soon as the $t$-aspect averaging is removed., Comment: 49 pages

Creative processes such as painting often involve creating different components of an image one by one. Can we build a computational model to perform this task? Prior works often fail by making global changes to the image, inserting objects in unrealistic spatial locations, and generating inaccurate lighting details. We observe that while state-of-the-art models perform poorly on object insertion, they can remove objects and erase the background in natural images very well. Inverting the direction of object removal, we obtain high-quality data for learning to insert objects that are spatially, physically, and optically consistent with the surroundings. With this scalable automatic data generation pipeline, we can create a dataset for learning object insertion, which is used to train our proposed text conditioned diffusion model. Qualitative and quantitative experiments have shown that our model achieves state-of-the-art results in object insertion, particularly for in-the-wild images. We show compelling results on diverse insertion prompts and images across various domains.In addition, we automate iterative insertion by combining our insertion model with beam search guided by CLIP.

Charge or spin-qubits can be realized by using gate-defined quantum dots (QDs) in semiconductors in a similar fashion to the processes used in CMOS for conventional field-effect transistors or more recent fin FET technology. However, to realize larger number of gate-defined qubits, multiples of gates with ultimately high resolution and fidelity is required. Electron beam lithography (EBL) offers flexible and tunable patterning of gate-defined spin-qubit devices for studying important quantum phenomena. While such devices are commonly realized by a positive resist process using metal lift-off, there are several clear limitations related to the resolution and the fidelity of patterning. Herein, we report a systematic study of an alternative TiN multi-gates definition approach based on the highest resolution hydrogen silsesquioxane (HSQ) EBL resist and all associated processing modes. The TiN gate arrays formed show excellent fidelity, dimensions down to 15 nm, various densities, and complexities. The processing modes developed were used to demonstrate applicability of this approach to forming multi-gate architectures for two types of spin-qubit devices prototypic to i) NW/fin-type FETs and ii) planar quantum well-type devices, both utilizing epi-grown Ge device layers on Si, where GeSn or Ge are the host materials for the QDs.

Let $S_l(M,N)$ denote a set of $\ell$ triples of positive integers whose parts have the same sum $M$ and the same product $N$. For each $2\leq\ell\leq 4$ we establish a connection between the set $S_l(M,N)$ and a family of elliptic curves. When $\ell=2$ and $3$, we use certain known parametrised sets $S_l(M,N)$ and respectively find families of elliptic curves of generic rank~$\geq 5$ and $\geq 6$, while for $\ell=4$ we first obtain a parametrised set $S_l(M,N)$, then construct a family of elliptic curves of generic rank $\geq 8$. Finally, we perform a computer search within these families to find specific curves with rank~$\geq 11$. The highest rank examples we found were two curves of rank~$14$., Comment: 11 pages, 4 tables

The exploitation of the strong light-matter coupling regime and exciton-polariton condensates has emerged as a compelling approach to introduce strong interactions and nonlinearities into numerous photonic applications, ranging from low-threshold topological lasers to ultrafast all-optical logic devices. The use of colloidal semiconductor quantum dots with strong three-dimensional confinement as the active material in these microcavities would be highly advantageous due to their versatile structural and compositional tunability and wet-chemical processability, as well as potentially enhanced, confinement-induced polaritonic interactions. Yet, to date, cavity exciton-polariton condensation has neither been achieved with epitaxial nor with colloidal quantum dots. Here, we demonstrate room-temperature polariton condensation in a thin film of monodisperse, colloidal CsPbBr$_3$ quantum dots placed in a tunable optical resonator with a Gaussian-shaped deformation serving as wavelength-scale potential well for the polaritons. The onset of polariton condensation under pulsed optical excitation is manifested in emission by its characteristic superlinear intensity dependence, reduced linewidth, blueshift, and extended temporal coherence. Our results, based on this highly engineerable class of perovskite materials with unique optical properties, pave the way for the development of polaritonic devices for ultrabright coherent light sources and photonic information processing.

This technical report introduces Docling, an easy to use, self-contained, MIT-licensed open-source package for PDF document conversion. It is powered by state-of-the-art specialized AI models for layout analysis (DocLayNet) and table structure recognition (TableFormer), and runs efficiently on commodity hardware in a small resource budget. The code interface allows for easy extensibility and addition of new features and models., Comment: Docling v1 report

The last decade has seen a renewed exploration of semiconductor materials for X-ray detection, foremost focusing on lead-based perovskites and other metal halides as direct-conversion materials and scintillators. However, the reported performance characteristics are often incomplete or misleading in assessing the practical utility of materials. This Perspective offers guidelines for choosing, estimating and presenting the relevant figures of merit. We also provide ready-to-used tools for calculating these figures of merit: MATLAB application, Mathcad worksheet and a website. The X-ray detectors for medical imaging are at focus for their increasing societal value and since they bring about the most stringent requirements as the image shall be acquired at as low as reasonably attainable (i.e. ALARA principle) dose received by the patient.

We analyze a combinatorial rule satisfied by the signs of principal minors of a real symmetric matrix. The sign patterns satisfying this rule are equivalent to uniform oriented Lagrangian matroids. We first discuss their structure and symmetries and then study their asymptotics, proving that almost all of them are not representable by real symmetric matrices. We offer several conjectures and experimental results concerning representable sign patterns and the topology of their representation spaces., Comment: 20 pages, 2 figures, 2 tables; code and data at https://mathrepo.mis.mpg.de/SymmetricPrincipalMinorSigns/ {\guillemotright} v4: Theorem 4.2 settles a conjecture in the previous version