Hasil untuk "Early Christian literature. Fathers of the Church, etc."

Andreas Varvarigos, Ali Maatouk, Jiasheng Zhang et al.

While large language models (LLMs) have become the de facto framework for literature-related tasks, they still struggle to function as domain-specific literature agents due to their inability to connect pieces of knowledge and reason across domain-specific contexts, terminologies, and nomenclatures. This challenge underscores the need for a tool that facilitates such domain-specific adaptation and enables rigorous benchmarking across literature tasks. To that end, we introduce LitBench, a benchmarking tool designed to enable the development and evaluation of domain-specific LLMs tailored to literature-related tasks. At its core, LitBench uses a data curation process that generates domain-specific literature sub-graphs and constructs training and evaluation datasets based on the textual attributes of the resulting nodes and edges. The tool is designed for flexibility, supporting the curation of literature graphs across any domain chosen by the user, whether high-level fields or specialized interdisciplinary areas. In addition to dataset curation, LitBench defines a comprehensive suite of literature tasks, ranging from node and edge level analyses to advanced applications such as related work generation. These tasks enable LLMs to internalize domain-specific knowledge and relationships embedded in the curated graph during training, while also supporting rigorous evaluation of model performance. Our results show that small domain-specific LLMs trained and evaluated on LitBench datasets achieve competitive performance compared to state-of-the-art models like GPT-4o and DeepSeek-R1. To enhance accessibility and ease of use, we open-source the tool along with an AI agent tool that streamlines data curation, model training, and evaluation.

David Brett, Anniek Myatt

Large language models (LLMs) have grown in their usage to provide support for question answering across numerous disciplines. The models on their own have already shown promise for answering basic questions, however fail quickly where expert domain knowledge is required or the question is nuanced. Scientific research often involves searching for relevant literature, distilling pertinent information from that literature and analysing how the findings support or contradict one another. The information is often encapsulated in the full text body of research articles, rather than just in the abstracts. Statements within these articles frequently require the wider article context to be fully understood. We have built an LLM-based system that performs such search and distillation of information encapsulated in scientific literature, and we evaluate our keyword based search and information distillation system against a set of biology related questions from previously released literature benchmarks. We demonstrate sparse retrieval methods exhibit results close to state of the art without the need for dense retrieval, with its associated infrastructure and complexity overhead. We also show how to increase the coverage of relevant documents for literature review generation.

Christian Grussler

Two nested classes of discrete-time linear time-invariant systems, which differ by the set of periodic signals that they leave invariant, are studied. The first class preserves the property of periodic monotonicity (period-wise unimodality). The second class is invariant to signals with at most two sign changes per period, and requires that periodic signals with zero sign changes are mapped to the same kind. Tractable characterizations for each system class are derived by the use and extension of total positivity theory via geometric interpretations. Central to our results is the characterization of sequentially convex contours via consecutive minors. Our characterizations also extend to the loop gain of Lur'e feedback systems as the considered signals sets are invariant under common static non-linearities, e.g., ideal relay, saturation, sigmoid function, quantizer, etc. The presented developments aim to form a base for future signal-based fixed-point theorems towards the prediction of self-sustained oscillations. Our examples on relay feedback systems indicate how periodic monotonicity preservation gives rise to useful insights towards this goal.

Marc Kamionkowski, Anubhav Mathur

Early dark energy solutions to the Hubble tension introduce an additional scalar field which is frozen at early times but becomes dynamical around matter-radiation equality. In order to alleviate the tension, the scalar's share of the total energy density must rapidly shrink from $\sim 10\%$ at the onset of matter domination to $\ll 1\%$ by recombination. This typically requires a steep potential that is imposed $\textit{ad hoc}$ rather than emerging from a concrete particle physics model. Here, we point out an alternative possibility: a homogeneous scalar field coupled quadratically to a cosmological background of light thermal relics (such as the Standard Model neutrino) will acquire an effective potential which can reproduce the dynamics necessary to alleviate the tension. We identify the relevant parameter space for this "thermo-coupled" scenario and study its unique phenomenology at the background level, including the back-reaction on the neutrino mass. Follow-up numerical work is necessary to determine the constraints placed on the model by early-time measurements.

Pierre Lamart, Yinan Yu, Christian Berger

Machine Learning (ML) is continuously permeating a growing amount of application domains. Generative AI such as Large Language Models (LLMs) also sees broad adoption to process multi-modal data such as text, images, audio, and video. While the trend is to use ever-larger datasets for training, managing this data efficiently has become a significant practical challenge in the industry-double as much data is certainly not double as good. Rather the opposite is important since getting an understanding of the inherent quality and diversity of the underlying data lakes is a growing challenge for application-specific ML as well as for fine-tuning foundation models. Furthermore, information retrieval (IR) from expanding data lakes is complicated by the temporal dimension inherent in time-series data which must be considered to determine its semantic value. This study focuses on the different semantic-aware techniques to extract embeddings from mono-modal, multi-modal, and cross-modal data to enhance IR capabilities in a growing data lake. Articles were collected to summarize information about the state-of-the-art techniques focusing on applications of embedding for three different categories of data modalities.

Guanyu Xu, Jiawei Hao, Li Shen et al.

Recently, the efficient deployment and acceleration of powerful vision transformers (ViTs) on resource-limited edge devices for providing multimedia services have become attractive tasks. Although early exiting is a feasible solution for accelerating inference, most works focus on convolutional neural networks (CNNs) and transformer models in natural language processing (NLP).Moreover, the direct application of early exiting methods to ViTs may result in substantial performance degradation. To tackle this challenge, we systematically investigate the efficacy of early exiting in ViTs and point out that the insufficient feature representations in shallow internal classifiers and the limited ability to capture target semantic information in deep internal classifiers restrict the performance of these methods. We then propose an early exiting framework for general ViTs termed LGViT, which incorporates heterogeneous exiting heads, namely, local perception head and global aggregation head, to achieve an efficiency-accuracy trade-off. In particular, we develop a novel two-stage training scheme, including end-to-end training and self-distillation with the backbone frozen to generate early exiting ViTs, which facilitates the fusion of global and local information extracted by the two types of heads. We conduct extensive experiments using three popular ViT backbones on three vision datasets. Results demonstrate that our LGViT can achieve competitive performance with approximately 1.8 $\times$ speed-up.

Marouane Daoui

This article conducts a literature review on the topic of monetary policy in developing countries and focuses on the effectiveness of monetary policy in promoting economic growth and the relationship between monetary policy and economic growth. The literature review finds that the activities of central banks in developing countries are often overlooked by economic models, but recent studies have shown that there are many factors that can affect the effectiveness of monetary policy in these countries. These factors include the profitability of central banks and monetary unions, the independence of central banks in their operations, and lags, rigidities, and disequilibrium analysis. The literature review also finds that studies on the topic have produced mixed results, with some studies finding that monetary policy has a limited or non-existent impact on economic growth and others finding that it plays a crucial role. The article aims to provide a comprehensive understanding of the current state of research in this field and to identify areas for future study.

Suchitra Narayanan, Jonathan P. Williams, John J. Tobin et al.

We present the first results from the Early Planet Formation in Embedded Disks (eDisk) ALMA Large Program toward Oph IRS43, a binary system of solar mass protostars. The 1.3 mm dust continuum observations resolve a compact disk, ~6au radius, around the northern component and show that the disk around the southern component is even smaller, <~3 au. CO, 13CO, and C18O maps reveal a large cavity in a low mass envelope that shows kinematic signatures of rotation and infall extending out to ~ 2000au. An expanding CO bubble centered on the extrapolated location of the source ~130 years ago suggests a recent outburst. Despite the small size of the disks, the overall picture is of a remarkably large and dynamically active region.

Vivek Kumar, Andreas Bauer, Christian Franz et al.

We report the magnetization, ac susceptibility, and specific heat of optically float-zoned single crystals of CePdAl$_{3}$. In comparison to the properties of polycrystalline CePdAl$_{3}$ reported in the literature, which displays a tetragonal crystal structure and no long-range magnetic order, our single crystals exhibit an orthorhombic structure ($Cmcm$) and order antiferromagnetically below a Néel temperature $T_{\rm N}$ = 5.6 K. The specific heat at zero-field shows a clear $λ$-type anomaly with a broad shoulder at $T_{\rm N}$. A conservative estimate of the Sommerfeld coefficient of the electronic specific heat, $γ= 121~\mathrm{mJ~K^{-2}~mol^{-1}}$, indicates a moderately enhanced heavy-fermion ground state. A twin microstructure evolves in the family of planes spanned by the basal plane lattice vectors $a_{\rm o}$ and $c_{\rm o}$, with the magnetic hard axis $b_{\rm o}$ common to all twins. The antiferromagnetic state is characterized by a strong magnetic anisotropy and a spin-flop transition induced under magnetic field along the easy direction, resulting in a complex magnetic phase diagram. Taken together our results reveal a high sensitivity of the magnetic and electronic properties of CePdAl$_{3}$ to its structural modifications.

Alex Buchel, Sera Cremonini, Laura Early

We set up a unified framework to efficiently compute the shear and bulk viscosities of strongly coupled gauge theories with gravitational holographic duals involving higher derivative corrections. We consider both Weyl$^4$ corrections, encoding the finite 't Hooft coupling corrections of the boundary theory, and Riemann$^2$ corrections, responsible for non-equal central charges $c\ne a$ of the theory at the ultraviolet fixed point. Our expressions for the viscosities in higher derivative holographic models are extracted from a radially conserved current and depend only on the horizon data.

James Young

The purpose of this review is to introduce the reader to graph kernels and the corresponding literature, with an emphasis on those with direct application to chemoinformatics. Graph kernels are functions that allow for the inference of properties of molecules and compounds, which can help with tasks such as finding suitable compounds in drug design. The use of kernel methods is but one particular way two quantify similarity between graphs. We restrict our discussion to this one method, although popular alternatives have emerged in recent years, most notably graph neural networks.

Steven J. Clark, Kyriakos Vattis, JiJi Fan et al.

An only early or only late time alteration to $Λ$CDM has been inadequate at resolving both the $H_0$ and $S_8$ tensions simultaneously; however, a combination of early and late time alterations to $Λ$CDM can provide a solution to both tensions. As an illustration, we examine a combined Early Dark Energy - Decaying Dark Matter model. While early dark energy has the ability to resolve the $H_0$ tension, it leads to a discrepancy in $S_8$ measurements. We show that the addition of decaying dark matter helps resolve the $S_8$ discrepancy that would otherwise be enhanced in an early dark energy model, while the latter is able to relieve the $H_0$ disagreement to within the 95th percentile interval. Our results show a preference for the combined model over $Λ$CDM with $Δ\rm{AIC} = -6.72$, hinting that both early and late universe modifications may be necessary to address the cosmological tensions.

Damien Christophe Jacques, Pierre Defourny

Early warning systems for food security rely on timely and accurate estimations of crop production. Several approaches have been developed to get early estimations of area and yield, the two components of crop production. The most common methods, based on Earth observation data, are image classification for crop area and correlation with vegetation index for crop yield. Regardless of the approach used, early estimators of cropland area, crop area or crop yield should have an accuracy providing lower production error than existing historical crop statistics. The objective of this study is to develop a methodological framework to define the accuracy requirements for early estimators of cropland area, crop area and crop yield in Senegal. These requirements are made according to (i) the inter-annual variability and the trend of historical data, (ii) the calendar of official statistics data collection, and (iii) the time at which early estimations of cropland area, crop area and crop yield can theoretically be available. This framework is applied to the seven main crops in Senegal using 20 years of crop production data. Results show that the inter-annual variability of crop yield is the main factor limiting the accuracy of pre-harvest production forecast. Estimators of cropland area can be used to improve production prediction of groundnuts, millet and rice, the three main crops in Senegal stressing the value of cropland mapping for food security. While applied to Senegal, this study could easily be reproduced in any country where reliable agricultural statistics are available.

Celina Rebello, Elaine Tavares

This article's objective is the identification of research opportunities in the current big data privacy domain, evaluating literature effects on secure informational assets. Until now, no study has analyzed such relation. Its results can foster science, technologies and businesses. To achieve these objectives, a big data privacy Systematic Literature Review (SLR) is performed on the main scientific peer reviewed journals in Scopus database. Bibliometrics and text mining analysis complement the SLR. This study provides support to big data privacy researchers on: most and least researched themes, research novelty, most cited works and authors, themes evolution through time and many others. In addition, TOPSIS and VIKOR ranks were developed to evaluate literature effects versus informational assets indicators. Secure Internet Servers (SIS) was chosen as decision criteria. Results show that big data privacy literature is strongly focused on computational aspects. However, individuals, societies, organizations and governments face a technological change that has just started to be investigated, with growing concerns on law and regulation aspects. TOPSIS and VIKOR Ranks differed in several positions and the only consistent country between literature and SIS adoption is the United States. Countries in the lowest ranking positions represent future research opportunities.

Christian Corda

We show that, based on important reasons, differently from the some recent claim in the literature, the YARK theory of gravity can reproduce neither the LIGO's "GW150914 signal" nor the other LIGO's detections of gravitational waves (GWs).

Maren Mahsereci, Lukas Balles, Christoph Lassner et al.

Early stopping is a widely used technique to prevent poor generalization performance when training an over-expressive model by means of gradient-based optimization. To find a good point to halt the optimizer, a common practice is to split the dataset into a training and a smaller validation set to obtain an ongoing estimate of the generalization performance. We propose a novel early stopping criterion based on fast-to-compute local statistics of the computed gradients and entirely removes the need for a held-out validation set. Our experiments show that this is a viable approach in the setting of least-squares and logistic regression, as well as neural networks.

Christian Bauckhage

We show that the objective function of conventional k-means clustering can be expressed as the Frobenius norm of the difference of a data matrix and a low rank approximation of that data matrix. In short, we show that k-means clustering is a matrix factorization problem. These notes are meant as a reference and intended to provide a guided tour towards a result that is often mentioned but seldom made explicit in the literature.

E. Alecian, O. Kochukhov, V. Petit et al.

To understand the origin of the magnetic fields in massive stars as well as their impact on stellar internal structure, evolution, and circumstellar environment, within the MiMeS project, we searched for magnetic objects among a large sample of massive stars, and build a sub-sample for in-depth follow-up studies required to test the models and theories of fossil field origins, magnetic wind confinement and magnetospheric properties, and magnetic star evolution. We obtained high-resolution spectropolarimetric observations of a large number of OB stars thanks to three large programs that have been allocated on the high-resolution spectropolarimeters ESPaDOnS, Narval, and the polarimetric module HARPSpol of the HARPS spectrograph. We report here on the methods and first analysis of the HARPSpol magnetic detections. We identified the magnetic stars using a multi-line analysis technique. Then, when possible, we monitored the new discoveries to derive their rotation periods, which are critical for follow-up and magnetic mapping studies. We also performed a first-look analysis of their spectra and identified obvious spectral anomalies (e.g., abundance peculiarities, Halpha emission), which are also of interest for future studies. In this paper, we focus on eight of the 11 stars in which we discovered or confirmed a magnetic field from the HARPSpol LP sample (the remaining three were published in a previous paper). Seven of the stars were detected in early-type Bp stars, while the last star was detected in the Ap companion of a normal early B-type star. We report obvious spectral and multiplicity properties, as well as our measurements of their longitudinal field strengths, and their rotation periods when we are able to derive them. We also discuss the presence or absence of Halpha emission with respect to the theory of centrifugally-supported magnetospheres. (Abriged)

Anne M. Green

Primordial Black Holes (PBHs) are, typically light, black holes which can form in the early Universe. There are a number of formation mechanisms, including the collapse of large density perturbations, cosmic string loops and bubble collisions. The number of PBHs formed is tightly constrained by the consequences of their evaporation and their lensing and dynamical effects. Therefore PBHs are a powerful probe of the physics of the early Universe, in particular models of inflation. They are also a potential cold dark matter candidate.

Andrew P. Billyard, Alan A. Coley, James E. Lidsey

We study early universe cosmologies derived from a scalar-tensor action containing cosmological constant terms and massless fields. The governing equations can be written as a dynamical system which contains no past or future asymptotic equilibrium states (i.e. no sources nor sinks). This leads to dynamics with very interesting mathematical behaviour such as the existence of heteroclinic cycles. The corresponding cosmologies have novel characteristics, including cyclical and bouncing behaviour possibly indicating chaos. We discuss the connection between these early universe cosmologies and those derived from the low-energy string effective action.