Search

Lopsided galaxies are late-type galaxies with a non-axisymmetric disc due to an uneven distribution of their stellar mass. Despite being a relatively common perturbation, several questions regarding its origin and the information that can be extracted from them about the evolutionary history of late-type galaxies. The advent of several large multi-band photometric surveys will allow us to statistically analyze this perturbation, with information that was not previously available. Given the strong correlation between lopsidedness and the structural properties of the galaxies, this paper aims to develop a method to automatically classify late-type galaxies between lopsided and symmetric. We seek to explore if an accurate classification can be obtain by only considering their internal properties, without additional information about the environment. We select 8000 late-type galaxies from TNG50. A Fourier decomposition of their stellar mass surface density is used to label galaxies as lopsided and symmetric. We trained a Random Forest classifier to rapidly and automatically identify this type of perturbations, exclusively using galaxies internal properties. We test different algorithms to deal with the imbalance of our data and select the most suitable approach based on the considered metrics. We show that our trained algorithm can provide a very accurate and rapid classification of lopsided galaxies. The excellent results obtained by our classifier strongly supports the hypothesis that lopsidedness is mainly a tracer of galaxies internal structures. We show that similar results can be obtained using observable quantities, readily obtainable from multi-bad photometric surveys. Our results show it allows a rapid and accurate classification of lopsided galaxies, allowing us to explore whether lopsidedness in present-day disc galaxies is connected to galaxies specific evolutionary histories., Comment: 15 pages, 10 figures. Submitted to A&A

Retrieval Augmented Generation (RAG) systems are a widespread application of Large Language Models (LLMs) in the industry. While many tools exist empowering developers to build their own systems, measuring their performance locally, with datasets reflective of the system's use cases, is a technological challenge. Solutions to this problem range from non-specific and cheap (most public datasets) to specific and costly (generating data from local documents). In this paper, we show that using public question and answer (Q&A) datasets to assess retrieval performance can lead to non-optimal systems design, and that common tools for RAG dataset generation can lead to unbalanced data. We propose solutions to these issues based on the characterization of RAG datasets through labels and through label-targeted data generation. Finally, we show that fine-tuned small LLMs can efficiently generate Q&A datasets. We believe that these observations are invaluable to the know-your-data step of RAG systems development., Comment: to be published in the 31st International Conference on Computational Linguistics (COLING 2025)

Current studies of large-scale asymmetries (i.e. lopsidedness) in the stellar density distribution of disk galaxies have mainly focused on the local Universe. Recent observations have found a significant fraction (over 60%) of lopsided galaxies at high-redshift ($1.5 < z < 3$), which is significantly larger than the fraction (~30%) observed in the nearby Universe. We aim to understand whether the more widespread lopsidedness at high- than low-redshift can be associated to environmental mechanisms being more effective in producing lopsided perturbations at high-redshift. At each redshift between $0 < z < 2$, we independently select a sample of disk-like galaxies from the IllustrisTNG simulations. We then characterize lopsidedness in the disks of galaxies at each redshift, study the relevant mechanisms generating lopsidedness, as well as the correlation between such perturbation, the local environment and the galaxy internal properties as a function of redshift. Consistent with previous and new observational results, we find that: 1) simulations predict a significant fraction (~60%) of lopsided galaxies at high-redshift ($1.5 < z < 2$), 2) the fraction of lopsided galaxies, as well as the lopsided amplitude, decreases from high- to low-redshift, and 3) there is not a significant dependence of lopsidedness on the local environment, but there is a strong correlation between the lopsided amplitude and basic galaxies' structural properties at all redshift between $0 < z < 2$. This means that, independent of the mechanisms on-setting lopsidedness, galaxies with low central stellar mass density and more extended disks are more susceptible of developing strong lopsidedness. We find that both recent interactions with mass-ratio >1:10 and gas accretion with subsequent star formation can produce lopsided perturbations at all redshift, but they are both significantly more effective at high-redshift., Comment: 15 pages, 11 figures. Submitted to A&A

Data preparation is the first and a very important step towards any Large Language Model (LLM) development. This paper introduces an easy-to-use, extensible, and scale-flexible open-source data preparation toolkit called Data Prep Kit (DPK). DPK is architected and designed to enable users to scale their data preparation to their needs. With DPK they can prepare data on a local machine or effortlessly scale to run on a cluster with thousands of CPU Cores. DPK comes with a highly scalable, yet extensible set of modules that transform natural language and code data. If the user needs additional transforms, they can be easily developed using extensive DPK support for transform creation. These modules can be used independently or pipelined to perform a series of operations. In this paper, we describe DPK architecture and show its performance from a small scale to a very large number of CPUs. The modules from DPK have been used for the preparation of Granite Models [1] [2]. We believe DPK is a valuable contribution to the AI community to easily prepare data to enhance the performance of their LLM models or to fine-tune models with Retrieval-Augmented Generation (RAG)., Comment: 10 pages, 7 figures

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

We determine the finite groups whose real irreducible representations have different degrees., Comment: 16 pages, comments welcome

Environment, Social, and Governance (ESG) KPIs assess an organization's performance on issues such as climate change, greenhouse gas emissions, water consumption, waste management, human rights, diversity, and policies. ESG reports convey this valuable quantitative information through tables. Unfortunately, extracting this information is difficult due to high variability in the table structure as well as content. We propose Statements, a novel domain agnostic data structure for extracting quantitative facts and related information. We propose translating tables to statements as a new supervised deep-learning universal information extraction task. We introduce SemTabNet - a dataset of over 100K annotated tables. Investigating a family of T5-based Statement Extraction Models, our best model generates statements which are 82% similar to the ground-truth (compared to baseline of 21%). We demonstrate the advantages of statements by applying our model to over 2700 tables from ESG reports. The homogeneous nature of statements permits exploratory data analysis on expansive information found in large collections of ESG reports., Comment: Accepted at the NLP4Climate workshop in the 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024)

Large language models (LLMs) trained on general domain corpora showed remarkable results on natural language processing (NLP) tasks. However, previous research demonstrated LLMs trained using domain-focused corpora perform better on specialized tasks. Inspired by this insight, we developed INDUS, a comprehensive suite of LLMs tailored for the closely-related domains of Earth science, biology, physics, heliophysics, planetary sciences and astrophysics, and trained using curated scientific corpora drawn from diverse data sources. The suite of models include: (1) an encoder model trained using domain-specific vocabulary and corpora to address NLP tasks, (2) a contrastive-learning based text embedding model trained using a diverse set of datasets to address information retrieval tasks and (3) smaller versions of these models created using knowledge distillation for applications which have latency or resource constraints. We also created three new scientific benchmark datasets, CLIMATE-CHANGE NER (entity-recognition), NASA-QA (extractive QA) and NASA-IR (IR) to accelerate research in these multi-disciplinary fields. We show that our models outperform both general-purpose (RoBERTa) and domain-specific (SCIBERT) encoders on these new tasks as well as existing tasks in the domains of interest. Furthermore, we demonstrate the use of these models in two industrial settings -- as a retrieval model for large-scale vector search applications and in automatic content tagging systems., Comment: EMNLP 2024 (Industry Track)

The structure of the character degree graphs $\Delta(G)$, i.e. the prime graphs on the set $\mathrm{cd}(G)$ of the irreducible character degrees of a finite group $G$, such that $G$ is solvable and $\Delta(G)$ has diameter three, remains an intriguing area of study. However, a comprehensive understanding of these structures remains elusive. In this paper, we prove some properties and provide an infinite series of examples of this class of graphs, building on the ideas of Mark Lewis.

The sharp Dirac cone of the electronic dispersion confers to graphene a remarkable sensitivity to strain. It is usually encoded in scalar and pseudo-vector potentials, induced by the modification of hopping parameters, which have given rise to new phenomena at the nanoscale such as giant pseudomagnetic fields and valley polarization. Here, we unveil the effect of these potentials on the quantum transport across a succession of strain-induced barriers. We use high-mobility, hBN-encapsulated graphene, transferred over a large (10x10 $\mu$m$^{2}$) crenellated hBN substrate. We show the emergence of a broad resistance ancillary peak at positive energy that arises from Klein tunneling barriers induced by the tensile strain at the trench edges. Our theoretical study, in quantitative agreement with experiment, highlights the balanced contributions of strain-induced scalar and pseudo-vector potentials on ballistic transport. Our results establish crenellated van der Waals heterostructures as a promising platform for strain engineering in view of applications and basic physics.

We present Deep Search DocQA. This application enables information extraction from documents via a question-answering conversational assistant. The system integrates several technologies from different AI disciplines consisting of document conversion to machine-readable format (via computer vision), finding relevant data (via natural language processing), and formulating an eloquent response (via large language models). Users can explore over 10,000 Environmental, Social, and Governance (ESG) disclosure reports from over 2000 corporations. The Deep Search platform can be accessed at: https://ds4sd.github.io., Comment: Accepted at the Demonstration Track of the 38th Annual AAAI Conference on Artificial Intelligence (AAAI 24)

We demonstrate, for the first time, Ramsey CPT spectroscopy with a miniature electro-optical bench associated to a dual-frequency generator based on combined optical injection locking and optical phase locking techniques Preliminary results show Ramsey CPT resonance with a contrast of 2% at the Cesium D2 line. Frequency difference locking loops lead to a contribution of Dick effect to fractional frequency stability lower than 1.7x10-13 at 1 s, in line with targeted clock stability of 5x10-13 at 1 s.

Large-scale asymmetries (i.e. lopsidedness) are a common feature in the stellar density distribution of nearby disk galaxies both in low- and high-density environments. In this work, we characterize the present-day lopsidedness in a sample of 1435 disk-like galaxies selected from the TNG50 simulation. We find that the percentage of lopsided galaxies (10%-30%) is in good agreement with observations if we use similar radial ranges to the observations. However, the percentage (58%) significantly increases if we extend our measurement to larger radii. We find a mild or lack of correlation between lopsidedness amplitude and environment at z=0 and a strong correlation between lopsidedness and galaxy morphology regardless of the environment. Present-day galaxies with more extended disks, flatter inner galactic regions and lower central stellar mass density (i.e. late-type disk galaxies) are typically more lopsided than galaxies with smaller disks, rounder inner galactic regions and higher central stellar mass density (i.e. early-type disk galaxies). Interestingly, we find that lopsided galaxies have, on average, a very distinct star formation history within the last 10 Gyr, with respect to their symmetric counterparts. Symmetric galaxies have typically assembled at early times (~8-6 Gyr ago) with relatively short and intense bursts of central star formation, while lopsided galaxies have assembled on longer timescales and with milder initial bursts of star formation, continuing building up their mass until z=0. Overall, these results indicate that lopsidedness in present-day disk galaxies is connected to the specific evolutionary histories of the galaxies that shaped their distinct internal properties., Comment: Submitted to MNRAS, 23 pages (including the Appendix), 23 figures

Transforming documents into machine-processable representations is a challenging task due to their complex structures and variability in formats. Recovering the layout structure and content from PDF files or scanned material has remained a key problem for decades. ICDAR has a long tradition in hosting competitions to benchmark the state-of-the-art and encourage the development of novel solutions to document layout understanding. In this report, we present the results of our \textit{ICDAR 2023 Competition on Robust Layout Segmentation in Corporate Documents}, which posed the challenge to accurately segment the page layout in a broad range of document styles and domains, including corporate reports, technical literature and patents. To raise the bar over previous competitions, we engineered a hard competition dataset and proposed the recent DocLayNet dataset for training. We recorded 45 team registrations and received official submissions from 21 teams. In the presented solutions, we recognize interesting combinations of recent computer vision models, data augmentation strategies and ensemble methods to achieve remarkable accuracy in the task we posed. A clear trend towards adoption of vision-transformer based methods is evident. The results demonstrate substantial progress towards achieving robust and highly generalizing methods for document layout understanding., Comment: ICDAR 2023, 10 pages, 4 figures

The function $\mathrm{P}_{\mathbf{v}}(G)$, measuring the proportion of the elements of a finite group $G$ that are zeros of irreducible characters of $G$, takes (as proved in [12]) only values $\frac{m-1}{m}$, for $1 \leq m \leq 6$, in the interval $[0, \mathrm{P}_{\mathbf{v}}(A_7))$.In this paper, we give a complete classification of the finite groups $G$ such that $\mathrm{P}_{\mathbf{v}}(G)=\frac{m-1}{m}$ for $m=1,2,\cdots ,6$., Comment: arXiv admin note: text overlap with arXiv:2201.00550

We present new, deep optical observations of the early-type galaxy NGC 4278, which is located in a small loose group. We find that the galaxy lacks fine substructure, i.e., it appears relaxed, out to a radius of $\sim$70 kpc. Our $g$- and $i$-band surface brightness profiles are uniform down to our deepest levels of $\sim$28 mag arcsec$^{-2}$. This spans an extremely large radial range of more than 14 half-mass radii. Combined with archival globular cluster (GC) number density maps and a new analysis of the total mass distribution obtained from archival Chandra X-ray data, we find that the red GC subpopulation traces well the stellar mass density profile from 2.4 out to even 14 half-mass radii, while the blue GC subpopulation traces the total mass density profile of the galaxy over a large radial range. Our results reinforce the scenario that red GCs form mostly in-situ along with the stellar component of the galaxy, while the blue GCs are more closely aligned with the total mass distribution in the halo and were accreted along with halo matter. We conclude that for galaxies where the X-ray emission from the hot halo is too faint to be properly observable and as such is not available to measure the dark matter profile, the blue GC population can be used to trace this dark matter component out to large radii., Comment: 11 pages, 10 figures, 1 table. Accepted for publication in MNRAS

Let $G$ be a finite group. Denoting by ${\rm{cd}}(G)$ the set of the degrees of the irreducible complex characters of $G$, we consider the {\it character degree graph} of $G$: this is the (simple, undirected) graph whose vertices are the prime divisors of the numbers in ${\rm{cd}}(G)$, and two distinct vertices $p$, $q$ are adjacent if and only if $pq$ divides some number in ${\rm{cd}}(G)$. This paper completes the classification, started in [5] and [6], of the finite non-solvable groups whose character degree graph has a {\it cut-vertex}, i.e. a vertex whose removal increases the number of connected components of the graph. More specifically, it was proved in [6] that these groups have a unique non-solvable composition factor $S$, and that $S$ is isomorphic to a group belonging to a restricted list of non-abelian simple groups. In [5] and [6] all isomorphism types for $S$ were treated, except the case \(S\cong{\rm{PSL}}_2(2^a)\) for some integer $a\geq 2$; the remaining case is addressed in the present paper., Comment: arXiv admin note: text overlap with arXiv:2208.03519

Foundation Models (FMs) have demonstrated unprecedented capabilities including zero-shot learning, high fidelity data synthesis, and out of domain generalization. However, as we show in this paper, FMs still have poor out-of-the-box performance on expert tasks (e.g. retrieval of car manuals technical illustrations from language queries), data for which is either unseen or belonging to a long-tail part of the data distribution of the huge datasets used for FM pre-training. This underlines the necessity to explicitly evaluate and finetune FMs on such expert tasks, arguably ones that appear the most in practical real-world applications. In this paper, we propose a first of its kind FETA benchmark built around the task of teaching FMs to understand technical documentation, via learning to match their graphical illustrations to corresponding language descriptions. Our FETA benchmark focuses on text-to-image and image-to-text retrieval in public car manuals and sales catalogue brochures. FETA is equipped with a procedure for completely automatic annotation extraction (code would be released upon acceptance), allowing easy extension of FETA to more documentation types and application domains in the future. Our automatic annotation leads to an automated performance metric shown to be consistent with metrics computed on human-curated annotations (also released). We provide multiple baselines and analysis of popular FMs on FETA leading to several interesting findings that we believe would be very valuable to the FM community, paving the way towards real-world application of FMs for practical expert tasks currently 'overlooked' by standard benchmarks focusing on common objects.

We investigate the origin of the large clustering signal detected in the angular distribution of the radio sources in the TGSS catalog. To do so, we cross-correlate the angular position of the radio sources with the Cosmic Microwave Background (CMB) lensing maps from the Planck satellite, since cross-correlation is expected to be insensitive to source of possible systematic errors that may generate a spurious clustering signal. The amplitude of the angular cross-correlation spectrum of TGSS-CMB lensing turns out to be much smaller than that of the TGSS auto-spectrum and consistent with that of the NVSS-CMB lensing cross spectrum. A result that confirms the spurious origin of the TGSS large scale clustering signal. We further compare the two cross-spectra with theoretical predictions that use various prescriptions from the literature, for the redshift counts of the radio sources, $N(z)$, and their bias $b(z)$. These models, that assume a $\Lambda$CDM cosmology and that were proposed to fit the NVSS auto-spectrum, fail to match the cross-spectra on large scale, though not by far. When the bias relation is let free to vary (model predictions are rather insensitive to the choice of the N(z)) the quality of the fit improves but a large bias ($ b_g = 2.53 \pm 0.11$) is required, which does not seem to be consistent with the observed clustering amplitude of the radio sources in the local universe. Whether this large cross-correlation amplitude represents a problem for the radio sources models, or for the $\Lambda$CDM framework itself, can only be clarified using next generation datasets featuring large number of objects. What our analysis does show is the possibility to remove the $N(z)$ and $b(z)$ degeneracy by combining angular and cross-correlation analyses., Comment: 21 pages, 18 figures

Let $G$ be a finite group, and let ${\rm{cd}}(G)$ denote the set of degrees of the irreducible complex characters of $G$. Define then the character degree graph $\Delta(G)$ as the (simple undirected) graph whose vertices are the prime divisors of the numbers in ${\rm{cd}}(G)$, and two distinct vertices $p$, $q$ are adjacent if and only if $pq$ divides some number in ${\rm{cd}}(G)$. This paper continues the work, started in [7], toward the classification of the finite non-solvable groups whose degree graph possesses a cut-vertex, i.e., a vertex whose removal increases the number of connected components of the graph. While, in [7], groups with no composition factors isomorphic to ${\rm{PSL}}_2(t^a)$ (for any prime power $t^a\geq 4$) were treated, here we consider the complementary situation in the case when $t$ is odd and $t^a> 5$. The proof of this classification will be then completed in the third and last paper of this series ([8]), that deals with the case $t=2$.

Accurate document layout analysis is a key requirement for high-quality PDF document conversion. With the recent availability of public, large ground-truth datasets such as PubLayNet and DocBank, deep-learning models have proven to be very effective at layout detection and segmentation. While these datasets are of adequate size to train such models, they severely lack in layout variability since they are sourced from scientific article repositories such as PubMed and arXiv only. Consequently, the accuracy of the layout segmentation drops significantly when these models are applied on more challenging and diverse layouts. In this paper, we present \textit{DocLayNet}, a new, publicly available, document-layout annotation dataset in COCO format. It contains 80863 manually annotated pages from diverse data sources to represent a wide variability in layouts. For each PDF page, the layout annotations provide labelled bounding-boxes with a choice of 11 distinct classes. DocLayNet also provides a subset of double- and triple-annotated pages to determine the inter-annotator agreement. In multiple experiments, we provide baseline accuracy scores (in mAP) for a set of popular object detection models. We also demonstrate that these models fall approximately 10\% behind the inter-annotator agreement. Furthermore, we provide evidence that DocLayNet is of sufficient size. Lastly, we compare models trained on PubLayNet, DocBank and DocLayNet, showing that layout predictions of the DocLayNet-trained models are more robust and thus the preferred choice for general-purpose document-layout analysis., Comment: 9 pages, 6 figures, 5 tables. Accepted paper at SIGKDD 2022 conference

Document understanding is a key business process in the data-driven economy since documents are central to knowledge discovery and business insights. Converting documents into a machine-processable format is a particular challenge here due to their huge variability in formats and complex structure. Accordingly, many algorithms and machine-learning methods emerged to solve particular tasks such as Optical Character Recognition (OCR), layout analysis, table-structure recovery, figure understanding, etc. We observe the adoption of such methods in document understanding solutions offered by all major cloud providers. Yet, publications outlining how such services are designed and optimized to scale in the cloud are scarce. In this paper, we focus on the case of document conversion to illustrate the particular challenges of scaling a complex data processing pipeline with a strong reliance on machine-learning methods on cloud infrastructure. Our key objective is to achieve high scalability and responsiveness for different workload profiles in a well-defined resource budget. We outline the requirements, design, and implementation choices of our document conversion service and reflect on the challenges we faced. Evidence for the scaling behavior and resource efficiency is provided for two alternative workload distribution strategies and deployment configurations. Our best-performing method achieves sustained throughput of over one million PDF pages per hour on 3072 CPU cores across 192 nodes., Comment: 11 pages, 7 figures, to be published in IEEE CLOUD 2022

Dual-comb LiDARs have the potential to perform high-resolution ranging at high speed. Here, through an implementation involving electro-optic modulators and heterodyne detection, we quantify the ranging systems trade-off between precision and non-ambiguity range (NAR) using a unique performance factor. We highlight the influence of the comb amplitude envelope on the precision with a distance measurement limited by the repetition rate of the optical comb. The influence of the combs repetition rate on the NAR and on the precision is illustrated through a setup allowing distance measurement with a tunable NAR. Finally, we demonstrate the impossibility to resolve different targets, quantify the impact on the measured distance and develop on the conditions in which non-linear effects of the interference make the measurement impossible., Comment: The authors acknowledge financial support by the Qombs Project (FET Flagship on Quantum Technologies grant No 820419) Quantum simulation and entanglement engineering in quantum cascade laser frequency combs

We study the present-day rotational velocity ($V_{rot}$) and velocity dispersion (${\sigma}$) profiles of the globular cluster (GC) systems in a sample of 50 lenticular (S0) galaxies from the E-MOSAICS galaxy formation simulations. We find that 82% of the galaxies have GCs that are rotating along the photometric major axis of the galaxy ($aligned$), while the remaining 18% of the galaxies do not ($misaligned$). This is generally consistent with the observations from the SLUGGS survey. For the $aligned$ galaxies, classified as $peaked$ $and$ $outwardly$ $decreasing$ (49%), $flat$ (24%) and $increasing$ (27%) based on the $V_{rot}/{\sigma}$ profiles out to large radii, we do not find any clear correlation between these present-day $V_{rot}/{\sigma}$ profiles of the GCs and the past merger histories of the S0 galaxies, unlike in previous simulations of galaxy stars. For just over half of the $misaligned$ galaxies, we find that the GC misalignment is the result of a major merger within the last 10 Gyr so that the $ex$-$situ$ GCs are misaligned by an angle between 0{\deg} (co-rotation) to 180{\deg} (counter-rotation) with respect to the $in$-$situ$ GCs, depending on the orbital configuration of the merging galaxies. For the remaining $misaligned$ galaxies, we suggest that the $in$-$situ$ metal-poor GCs, formed at early times, have undergone more frequent kinematic perturbations than the $in$-$situ$ metal-rich GCs. We also find that the GCs accreted early and the $in$-$situ$ GCs are predominantly located within 0.2 virial radii ($R_{200}$) from the centre of galaxies in 3D phase-space diagrams., Comment: 19 pages, 13 figures

We prove that the function $\mathrm{P}_{\mathrm{v}}(G)$, measuring the proportion of the elements of a finite group $G$ that are zeros of irreducible characters of $G$, takes very sparse values in a large segment of the $[0,1]$ interval.

Despite their ubiquity throughout the Universe, the formation of S0 galaxies remains uncertain. Recent observations have revealed that S0 galaxies make up a diverse population which is difficult to explain with a single formation pathway, suggesting that the picture of how these galaxies form is more complicated than originally envisioned. Here we take advantage of the latest hydrodynamical cosmological simulations and follow up these studies with an investigation into the formation histories of S0s in IllustrisTNG. We first classify IllustrisTNG galaxies in a way which is fully consistent with the observations, and reproduce the observed photometric and environmental distributions seen for the S0 population. We then trace the formation histories of S0 galaxies back through time, identifying two main distinct pathways; those which experienced gas stripping via group infalls (37 percent of S0s) or significant merger events (57 percent). We find that those forming via mergers feature a transient star-forming ring, whose present-day occurrence rate matches observations. We find that these formation pathways together can reproduce the range in rotational support in observed S0s, concluding that there are two main formation pathways for S0 galaxies., Comment: 17 pages, 16 figures, accepted for publication in MNRAS

We investigate the kinematic properties of nine nearby early-type galaxies with evidence of a disk-like component. Three of these galaxies are located in the field, five in the group and only one in the cluster environment. By combining the kinematics of the stars with those of the globular clusters (GCs) and planetary nebulae (PNe), we probe the outer regions of our galaxies out to $\sim$4-6 Re. Six galaxies have PNe and red GCs that show good kinematic alignment with the stars, whose rotation occurs along the photometric major-axis of the galaxies, suggesting that both the PNe and red GCs are good tracers of the underlying stellar population beyond that traced by the stars. Additionally, the blue GCs also show rotation that is overall consistent with that of the red GCs in these six galaxies. The remaining three galaxies show kinematic twists and misalignment of the PNe and GCs with respect to the underlying stars, suggesting recent galaxy interactions. From the comparison with simulations, we propose that all six aligned galaxies that show similar dispersion-dominated kinematics at large radii (>2-3 Re) had similar late ($z<1$) assembly histories characterised by mini mergers (mass-ratio <1:10). The different Vrot/$\sigma$ profiles are then the result of an early ($z>1$) minor merger (1:10< mass-ratio <1:4) for the four galaxies with peaked and decreasing Vrot/$\sigma$ profiles and of a late minor merger for the two galaxies with flat Vrot/$\sigma$ profiles. The three mis-aligned galaxies likely formed through multiple late minor mergers that enhanced their velocity dispersion at all radii, or a late major merger that spun-up both the GC sub-populations at large radii. Therefore, lenticular galaxies can have complex merger histories that shape their characteristic kinematic profile shapes., Comment: 31 pages, 12 figures. Accepted for publication in MNRAS on 2021 April 09

For an irreducible complex character $\chi$ of the finite group $G$, let $\pi(\chi)$ denote the set of prime divisors of the degree $\chi(1)$ of $\chi$. Denote then by $\rho(G)$ the union of all the sets $\pi(\chi)$ and by $\sigma(G)$ the largest value of $|\pi(\chi)|$, as $\chi$ runs in ${\rm{Irr}}(G)$. The $\rho$-$\sigma$ conjecture, formulated by Bertram Huppert in the 80's, predicts that $|\rho(G)|\leq 3\sigma(G)$ always holds, whereas $|\rho(G)|\leq 2\sigma(G)$ holds if $G$ is solvable; moreover, O. Manz and T.R. Wolf proposed a "strengthened" form of the conjecture in the general case, asking whether $|\rho(G)|\leq 2\sigma(G)+1$ is true for every finite group $G$. In this paper we study the strengthened $\rho$-$\sigma$ conjecture for the class of finite groups having a trivial Fitting subgroup: in this context, we prove that the conjecture is true provided $\sigma(G)\leq 5$, but it is false in general if $\sigma(G)\geq 6$. Instead, we establish that $|\rho(G)|\leq 3\sigma(G)-4$ holds for every finite group with a trivial Fitting subgroup and with $\sigma(G)\geq 6$ (this being the right, best possible bound). Also, we improve the up-to-date best bound for the solvable case, showing that we have $|\rho(G)|\leq 3\sigma(G)$ whenever $G$ belongs to one particular class including all the finite solvable groups.

The number of published PDF documents has increased exponentially in recent decades. There is a growing need to make their rich content discoverable to information retrieval tools. In this paper, we present a novel approach to document structure recovery in PDF using recurrent neural networks to process the low-level PDF data representation directly, instead of relying on a visual re-interpretation of the rendered PDF page, as has been proposed in previous literature. We demonstrate how a sequence of PDF printing commands can be used as input into a neural network and how the network can learn to classify each printing command according to its structural function in the page. This approach has three advantages: First, it can distinguish among more fine-grained labels (typically 10-20 labels as opposed to 1-5 with visual methods), which results in a more accurate and detailed document structure resolution. Second, it can take into account the text flow across pages more naturally compared to visual methods because it can concatenate the printing commands of sequential pages. Last, our proposed method needs less memory and it is computationally less expensive than visual methods. This allows us to deploy such models in production environments at a much lower cost. Through extensive architectural search in combination with advanced feature engineering, we were able to implement a model that yields a weighted average F1 score of 97% across 17 distinct structural labels. The best model we achieved is currently served in production environments on our Corpus Conversion Service (CCS), which was presented at KDD18 (arXiv:1806.02284). This model enhances the capabilities of CCS significantly, as it eliminates the need for human annotated label ground-truth for every unseen document layout. This proved particularly useful when applied to a huge corpus of PDF articles related to COVID-19., Comment: 9 pages, 2 tables, 4 figures, uses aaai21.sty. Accepted at the "Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21)". Received the "IAAI-21 Innovative Application Award"

The mechanical stability of commercial GaP/Si templates during thermal an-nealing and subsequent MOCVD growth of GaP and AlGaP is investigated.Although the GaP layer of the template originally presents an excellent surfacemorphology, annealing at high enough temperatures to remove the native oxideprior to growth leads to plastic relaxation, accompanied by a variety of defects,including a dense grid of micro-twins. These micro-twins detrimentally affectGaP and AlGaP layers grown subsequently on the template., Comment: 6 figures