Zahra Khalila, Arbi Haza Nasution, Winda Monika
et al.
Accurate and contextually faithful responses are critical when applying large language models (LLMs) to sensitive and domain-specific tasks, such as answering queries related to quranic studies. General-purpose LLMs often struggle with hallucinations, where generated responses deviate from authoritative sources, raising concerns about their reliability in religious contexts. This challenge highlights the need for systems that can integrate domain-specific knowledge while maintaining response accuracy, relevance, and faithfulness. In this study, we investigate 13 open-source LLMs categorized into large (e.g., Llama3:70b, Gemma2:27b, QwQ:32b), medium (e.g., Gemma2:9b, Llama3:8b), and small (e.g., Llama3.2:3b, Phi3:3.8b). A Retrieval-Augmented Generation (RAG) is used to make up for the problems that come with using separate models. This research utilizes a descriptive dataset of Quranic surahs including the meanings, historical context, and qualities of the 114 surahs, allowing the model to gather relevant knowledge before responding. The models are evaluated using three key metrics set by human evaluators: context relevance, answer faithfulness, and answer relevance. The findings reveal that large models consistently outperform smaller models in capturing query semantics and producing accurate, contextually grounded responses. The Llama3.2:3b model, even though it is considered small, does very well on faithfulness (4.619) and relevance (4.857), showing the promise of smaller architectures that have been well optimized. This article examines the trade-offs between model size, computational efficiency, and response quality while using LLMs in domain-specific applications.
Federico Lavorenti, Elizabeth A. Jensen, Sae Aizawa
et al.
Mercury is the closest planet to the Sun, possesses a weak intrinsic magnetic field and has only a very tenuous atmosphere (exosphere). These three conditions result in a direct coupling between the plasma emitted from the Sun (namely the solar wind) and Mercury's surface. The planet's magnetic field leads to a non-trivial pattern of plasma precipitation onto the surface, that is expected to contribute to the alteration of the regolith over geological time scales. The goal of this work is to study the solar wind plasma precipitation onto the surface of Mercury from a geographical perspective, as opposed to the local-time-of-day approach of previous precipitation modeling studies. We employ solar wind precipitation maps for protons and electrons from two fully-kinetic numerical simulations of Mercury's plasma environment. These maps are then integrated over two full Mercury orbits (176 Earth days). We found that the plasma precipitation pattern at the surface is most strongly affected by the upstream solar wind conditions, particularly by the interplanetary magnetic field direction, and less by Mercury's 3:2 spin-orbit resonance. We also found that Mercury's magnetic field is able to shield the surface from roughly 90% of the incoming solar wind flux. At the surface, protons have a broad energy distribution from below 500 eV to more than 1.5 keV; while electrons are mostly found in the range 0.1-4 keV. These results will help to better constrain space weathering and exosphere source processes at Mercury, as well as to interpret observations by the ongoing ESA/JAXA BepiColombo mission.
Automated decision-making systems are becoming increasingly ubiquitous, which creates an immediate need for their interpretability and explainability. However, it remains unclear whether users know what insights an explanation offers and, more importantly, what information it lacks. To answer this question we conducted an online study with 200 participants, which allowed us to assess explainees' ability to realise explicated information -- i.e., factual insights conveyed by an explanation -- and unspecified information -- i.e, insights that are not communicated by an explanation -- across four representative explanation types: model architecture, decision surface visualisation, counterfactual explainability and feature importance. Our findings uncover that highly comprehensible explanations, e.g., feature importance and decision surface visualisation, are exceptionally susceptible to misinterpretation since users tend to infer spurious information that is outside of the scope of these explanations. Additionally, while the users gauge their confidence accurately with respect to the information explicated by these explanations, they tend to be overconfident when misinterpreting the explanations. Our work demonstrates that human comprehension can be a double-edged sword since highly accessible explanations may convince users of their truthfulness while possibly leading to various misinterpretations at the same time. Machine learning explanations should therefore carefully navigate the complex relation between their full scope and limitations to maximise understanding and curb misinterpretation.
Wei-Chih Hung, Vincent Casser, Henrik Kretzschmar
et al.
The 3D Average Precision (3D AP) relies on the intersection over union between predictions and ground truth objects. However, camera-only detectors have limited depth accuracy, which may cause otherwise reasonable predictions that suffer from such longitudinal localization errors to be treated as false positives. We therefore propose variants of the 3D AP metric to be more permissive with respect to depth estimation errors. Specifically, our novel longitudinal error tolerant metrics, LET-3D-AP and LET-3D-APL, allow longitudinal localization errors of the prediction boxes up to a given tolerance. To evaluate the proposed metrics, we also construct a new test set for the Waymo Open Dataset, tailored to camera-only 3D detection methods. Surprisingly, we find that state-of-the-art camera-based detectors can outperform popular LiDAR-based detectors with our new metrics past at 10% depth error tolerance, suggesting that existing camera-based detectors already have the potential to surpass LiDAR-based detectors in downstream applications. We believe the proposed metrics and the new benchmark dataset will facilitate advances in the field of camera-only 3D detection by providing more informative signals that can better indicate the system-level performance.
Digital audio processing tools offer music researchers the opportunity to examine both non-notated music and music as performance. This chapter summarises the types of information that can be extracted from audio as well as currently available audio tools for music corpus studies. The survey of extraction methods includes both a primer on signal processing and background theory on audio feature extraction. The survey of audio tools focuses on widely used tools, including both those with a graphical user interface, namely Audacity and Sonic Visualiser, and code-based tools written in the C/C++, Java, MATLAB, and Python computer programming languages.
Intensity diffraction tomography (IDT) refers to a class of optical microscopy techniques for imaging the 3D refractive index (RI) distribution of a sample from a set of 2D intensity-only measurements. The reconstruction of artifact-free RI maps is a fundamental challenge in IDT due to the loss of phase information and the missing cone problem. Neural fields (NF) has recently emerged as a new deep learning (DL) approach for learning continuous representations of physical fields. NF uses a coordinate-based neural network to represent the field by mapping the spatial coordinates to the corresponding physical quantities, in our case the complex-valued refractive index values. We present DeCAF as the first NF-based IDT method that can learn a high-quality continuous representation of a RI volume from its intensity-only and limited-angle measurements. The representation in DeCAF is learned directly from the measurements of the test sample by using the IDT forward model, without any ground-truth RI maps. We qualitatively and quantitatively evaluate DeCAF on the simulated and experimental biological samples. Our results show that DeCAF can generate high-contrast and artifact-free RI maps and lead to up to 2.1 times reduction in MSE over existing methods.
Mung bean (Vigna radiata L.) is an important but understudied food legume in Asia and now worldwide. Genetic studies may help to accelerate the exploitation of new genes for breeding in this crop. Here, we used a recombination inbred line population to construct an SNP genetic linkage map by genome sequencing technology. We obtained 21,508 high-quality SNP markers integrated into 1,946 bin markers that were mapped onto 11 linkage groups (LGs) with 99–258 bin markers per LG. The total genetic length of the map was 1060.2 cM (38.76–168.03 cM per LG), with an average distance between markers of 0.54 cM. However, there were 18 gaps >5 cM, distribution on LG1, 3, 5, 7, and 9. Gene mapping for lobed and indented leaflets was conducted using the map. A major quantitative trait locus (QTL) associated with indented leaflets was detected on chromosome 10, with phenotypic variation explained (PVE) values of 39.7% and 45.4% under two different environments. Several QTLs for lobed leaflets were detected and most of them were tightly linked together on Chromosome 3. However, only one major QTL, which explained the largest phenotypic variation (27.7–69.5%), was stably detected under two different environments using both R and Q methods. In the two main stable QTLs regions on chromosomes 3 and 10, candidate genes for regulating the molecular mechanism of different leaflet shapes were detected by functional annotation. The overlap of major QTLs under different environments indicated that the present map would be good enough for precisely mapping genes, and both the QTL analysis and gene prediction were useful for investigating the mechanism of leaf development in mung bean or legumes.
Early education is critical for improving energy efficiency. The purpose of this study is to explore the feasibility of Interactive Cross-Border Classes to increase awareness of energy efficiency among middle school students. We designed and tested an Interactive Cross-Border class between Chilean and Peruvian 8th-grade classes. The classes were synchronously connected and all students answered open-ended questions on an online platform. Some of the questions were designed to check conceptual understanding while others asked for suggestions of how to develop their economies while keeping CO₂ air concentration at acceptable levels. In real-time, the teacher reviewed the students’ written answers and the concept maps that were automatically generated based on their responses. Students peer-reviewed their classmates’ suggestions. This is part of an Asia-Pacific Economic Cooperation (APEC) STEM Education project on Energy Efficiency using APEC databases. We found high levels of student engagement, where students discussed not only the cross-cutting nature of energy, but also its relation to socioeconomic development and CO₂ emissions, and the need to work together to improve energy efficiency. In conclusion, Interactive Cross-Border classes are a feasible educational alternative, with potential as a scalable public policy strategy for improving awareness of energy efficiency among the population.
Land cover is a commonly used index for characterizing land surfaces and the corresponding human or natural processes. Remote sensing-based land cover maps of different times can be compared to extract changes in land cover. However, remote sensing images reflect only the instantaneous state of the Earth’s surface. Spectral characteristics cannot correctly reflect the actual state due to the differences caused by seasonal changes, and this inability results in the limited classification accuracy of land cover products. Spurious changes will occur when products of different time phases are compared. A method of spurious change detection based on a geo-eco zoning rule base is proposed in this study to solve this problem. The structure and content of the rule base are designed, and a rule base management system is established to identify spurious changes. Land cover products of Southeast Asia are selected as the experimental area to verify the accuracy of the change patches after eliminating spurious changes. Results show that the accuracy of change detection is improved by using the rule base of geo-eco zoning to identify spurious changes.
2 sitasi
en
Computer Science, Environmental Science
We study how a single paper affects the Impact Factor (IF) by analyzing data from 3,088,511 papers published in 11639 journals in the 2017 Journal Citation Reports of Clarivate Analytics. We find that IFs are highly volatile. For example, the top-cited paper of 381 journals caused their IF to increase by more than 0.5 points, while for 818 journals the relative increase exceeded 25%. And one in 10 journals had their IF boosted by more than 50% by their top three cited papers. Because the single-paper effect on the IF is inversely proportional to journal size, small journals are rewarded much more strongly than large journals for a highly-cited paper, while they are penalized more for a low-cited paper, especially if their IF is high. This skewed reward mechanism incentivizes high-IF journals to stay small, to remain competitive in rankings. We discuss the implications for breakthrough papers appearing in prestigious journals. We question the reliability of IF rankings given the high IF sensitivity to a few papers for thousands of journals.
In this note, we study the directed first passage percolation introduced in [F. Comets, R. Fukushima, S. Nakajima and N. Yoshida: Journal of Statistical Physics, 161-(3), 577-597 (2015)]. It is proved that the shortest path from the origin to the $n$-th hyperplane makes a jump larger than a small power of $\log n$. Some numerical results are also provided, which indicates that the maximal jump size is much larger in a certain parameter region.
Let $G$ be a subgroup of ${\rm PGL}_2({\mathbb F}_q)$, where $q$ is any prime power, and let $Q \in {\mathbb F}_q[x]$ such that ${\mathbb F}_q(x)/{\mathbb F}_q(Q(x))$ is a Galois extension with group $G$. By explicitly computing the Artin map on unramified degree-1 primes in ${\mathbb F}_q(Q)$ for various groups $G$, interesting new results emerge about finite fields, additive polynomials, and conjugacy classes of ${\rm PGL}_2({\mathbb F}_q)$. For example, by taking $G$ to be a unipotent group, one obtains a new characterization for when an additive polynomial splits completely over ${\mathbb F}_q$. When $G = {\rm PGL}_2({\mathbb F}_q)$, one obtains information about conjugacy classes of ${\rm PGL}_2({\mathbb F}_q)$. When $G$ is the group of order 3 generated by $x \mapsto 1 - 1/x$, one obtains a natural tripartite symbol on ${\mathbb F}_q$ with values in ${\mathbb Z}/3{\mathbb Z}$. Some of these results generalize to ${\rm PGL}_2(K)$ for arbitrary fields $K$. Apart from the introduction, this article is written from first principles, with the aim to be accessible to graduate students or advanced undergraduates. An earlier draft of this article was published on the Math arXiv in June 2019 under the title {\it More structure theorems for finite fields}.
Within Plasmodium falciparum merozoite surface protein 1 (MSP1), the N-terminal block 2 region is a highly polymorphic target of naturally acquired antibody responses. The antigenic diversity is determined by complex repeat sequences as well as non-repeat sequences, grouping into three major allelic types that appear to be maintained within populations by natural selection. Within these major types, many distinct allelic sequences have been described in different studies, but the extent and significance of the diversity remains unresolved. To survey the diversity more extensively, block 2 allelic sequences in the msp1 gene were characterized in 2400 P. falciparum infection isolates with whole genome short read sequence data available from the Pf3K project, and compared with the data from previous studies. Mapping the short read sequence data in the 2400 isolates to a reference library of msp1 block 2 allelic sequences yielded 3815 allele scores at the level of major allelic family types, with 46% of isolates containing two or more of these major types. Overall frequencies were similar to those previously reported in other samples with different methods, the K1-like allelic type being most common in Africa, MAD20-like most common in Southeast Asia, and RO33-like being the third most abundant type in each continent. The rare MR type, formed by recombination between MAD20-like and RO33-like alleles, was only seen in Africa and very rarely in the Indian subcontinent but not in Southeast Asia. A combination of mapped short read assembly approaches enabled 1522 complete msp1 block 2 sequences to be determined, among which there were 363 different allele sequences, of which 246 have not been described previously. In these data, the K1-like msp1 block 2 alleles are most diverse and encode 225 distinct amino acid sequences, compared with 123 different MAD20-like, 9 RO33-like and 6 MR type sequences. Within each of the major types, the different allelic sequences show highly skewed geographical distributions, with most of the more common sequences being detected in either Africa or Asia, but not in both. Allelic sequences of this extremely polymorphic locus have been derived from whole genome short read sequence data by mapping to a reference library followed by assembly of mapped reads. The catalogue of sequence variation has been greatly expanded, so that there are now more than 500 different msp1 block 2 allelic sequences described. This provides an extensive reference for molecular epidemiological genotyping and sequencing studies, and potentially for design of a multi-allelic vaccine.
Katarzyna Biesialska, Xavier Franch, Victor Muntés-Mulero
Conducting empirical research in software engineering industry is a process, and as such, it should be generalizable. The aim of this paper is to discuss how academic researchers may address some of the challenges they encounter during conducting empirical research in the software industry by means of a systematic and structured approach. The protocol developed in this paper should serve as a practical guide for researchers and help them with conducting empirical research in this complex environment.
The Comparative Study of Tectonic Environment and Metallogenic Laws of Cu–Au Mineralization in the Indo-China Peninsula and Western Pacific Giant metallogenic belt takes the Indo-China Peninsula and the giant Western Pacific metallogenic belt as the study area and targets Cu and Au minerals. Data was systematically collected via international cooperation based on the compilation of 1:1,500,000 scale Geological and Mineral Map of Indo-China Peninsula and Western Pacific Giant metallogenic belt, Southeast Asia. Study results were deepened and maps in English and Chinese versions were published after studying the metallogenic environment, ore-controlling factors and metallogenic laws of typical deposits in the study area covering Laos, Thailand, Burma, Papua New Guinea, etc. and combining them with ETM+, Hyperion and ASTER multi-source remote sensing images. Cu–Au deposits mainly in Laos were analyzed, while prospecting demonstration and drilling verification were conducted in the Aven deposit area, Laos. The study achieved positive prospecting effects. Moreover, using MapGIS and ArcGIS technical platforms, maps of metallogenic laws of mineralization zones/regions were compiled, and quantitative evaluation models for different minerals and a spatial dataset of Au–Cu deposits in significant metallogenic zones/regions in Southeast Asia were developed. In addition, using the mathematical modeling method, resource potential assessment was conducted for the study area, and mineralization prospects for different minerals were delineated within the study area. This dataset is not only a summary of mineral resource zoning and resource potential assessment, but also a theoretical basis for scientifically guiding China’s prospecting deployment in other countries (the Belt and Road Initiative).
Saffron (Crocus sativus) has been one of the most important spices in the world since ancient times. Though there is a variety of information about its origin, it is believed to have emerged in Iran, Turkey, or Greece and have spread across the world. Saffron, which is commonly produced in the Mediterranean and Southwest Asian countries, is used in many different fields such as painting, medicine, perfumery, and food. Climate and soil conditions also play an important role in saffron production. Saffron, which shows a flexible characteristic in terms of temperature demand, shows resistance to cold temperatures down to -18 degrees and to summer temperatures up to 45 degrees. The type of soil suggested for the development of the plant is clay loam soil. Although saffron growing in Turkey has lost its former significance today, it is still carried out on a small scale in Safranbolu city. This study intends to show the positive effects of climatic elements prevailing in the region on the production of saffron, the world's most expensive plant, and to reestablish its former importance in the region. The climate data used in the present study were taken from the General Directorate of Meteorology of Turkey. The climatic conditions required by the Thornthwaite climate classification method. Systematic approach was used as a research method. The maps used in the study were prepared on ArcGIS 10.3 GIS (Geography Information System) package. According to the Thornthwaite method, Safranbolu is arid-low humidity in terms of rainfall activity, has a 2nd level mesothermal climate, has no water excess, and is closed to sea effect. Considering the climate demands of saffron, it can be said that the climate of Safranbolu city is suitable for saffron growing. In Safranbolu, saffron plants are currently cultivated only in few villages. However, more effort should be made to enhance the production of saffron, which is as precious as gold.