Large Language Models (LLMs) have emerged as powerful tools in the domain of artificial intelligence, specifically enhancing natural language processing capabilities. These models leverage extensive datasets and sophisticated algorithms to automate and enhance tasks traditionally performed by humans, thereby revolutionizing innovation and entrepreneurship. LLMs accelerate product development, streamline business operations, and enable precise and rapid decision-making, all of which are crucial for maintaining competitiveness in dynamic markets. This paper categorizes the applications of LLMs in innovation and entrepreneurship into three main areas: technological innovation, strategic decision-making, and process automation. By exploring various integrations of LLMs in entrepreneurial ventures, this work offers both theoretical insights and practical examples, underscoring the transformative impact of LLMs in shaping modern business landscapes.

LLMs are a compressed repository of a vast corpus of valuable information on which they are trained. Therefore, this work hypothesizes that LLMs such as Llama, Orca, Falcon, and Mistral can be used for misinformation detection by making them cross-check new information with the repository on which they are trained. Accordingly, this paper describes the findings from the investigation of the abilities of LLMs in detecting misinformation on multiple datasets. The results are interpreted using explainable AI techniques such as LIME, SHAP, and Integrated Gradients. The LLMs themselves are also asked to explain their classification. These complementary approaches aid in better understanding the inner workings of misinformation detection using LLMs and lead to conclusions about their effectiveness at the task.

This survey provides a comprehensive overview of in-context learning (ICL) in large language models (LLMs), a phenomenon where models can adapt to new tasks without parameter updates by leveraging task-relevant information within the input context. We explore the definition and mechanisms of ICL, investigate the factors contributing to its emergence, and discuss strategies for optimizing and effectively utilizing ICL in various applications. Through a systematic review of recent literature, we first clarify what ICL is, distinguishing it from traditional fine-tuning approaches and highlighting its unique characteristics. We then delve into the underlying causes of ICL, examining theories ranging from implicit meta-learning during pre-training to the emergence of task vectors in LLMs. The survey also covers various approaches to enhance ICL performance, including prompt engineering techniques, demonstration selection strategies, and methods for improving generalization across diverse tasks. Additionally, we discuss the limitations and challenges of ICL, such as its sensitivity to demonstration ordering and potential biases. By synthesizing findings from numerous studies, we aim to provide researchers and practitioners with a clear understanding of the current state of ICL research, its practical implications, and promising directions for future investigation. This survey serves as a valuable resource for those seeking to leverage ICL capabilities in LLMs and contributes to the ongoing discourse on the remarkable adaptability of these models.

Large language models have recently made significant progress in natural language processing, and there is observation that these models exhibit reasoning abilities when they are sufficiently large. This has sparked considerable research interest since reasoning is a hallmark of human intelligence that is widely considered missed in artificial intelligence systems. Due to large size of these models, evaluation of LLMs’ reasoning ability is largely empirical. Creating datasets to evaluate the reasoning ability of LLMs is therefore an important area of LLM research. A key open question is whether LLMs reason or simply recite memorized texts they have encountered during their training phase. This work conducts simple experiments using Cheryl’s Birthday Puzzle and Cheryl’s Age Puzzle and their variants created in this work to investigate whether LLMs recite or reason and discovers that LLMs tend to recite memorized answers for well-known questions, which appear more frequently on the internet, even though such answers may not be sensible for the modified versions of the questions. When presented with less well-known questions, it is observed that LLMs answer with correct reasoning more frequently. A possible inference is that LLMs tend to reason on less well-known questions but recite memorized answers on popular questions. As a result, to accurately evaluate the reasoning ability of LLMs, it is essential to create new datasets to ensure that LLMs are elicited to truly use their reasoning ability to generate responses to the presented questions. In view of the finding, this work proposes a new dataset comprising of unseen questions requiring semantic and deductive logical reasoning skills to elicit reasoning ability from LLMs. The proposed evaluation framework is based on the intuition that some questions or puzzles can only be answered through the mastery of reasoning needed for situational awareness, and the dataset consists of 84 logical reasoning questions on room assignment subject to constraints. The proposed evaluation framework has several desirable properties, including resilience to training data contamination, ease of response verification, extensibility, usefulness and automated test case generation. This work then applies the proposed dataset to evaluate the reasoning ability of state-of-the-art LLMs, including GPT-3, GPT-4, Llama-3.1, Gemini-1.5 and Claude-3.5, and compare their performance with the performance of human intelligence on the dataset.

Although learning analytics (LA) holds great potential to improve teaching and learning, LA research and practice are currently riddled with limitations that impact every stage of the LA life cycle. The present paper offers an overview of these challenges before proposing strategies to overcome them and exploring how the recent innovations brought forth by language models can improve LA research and practice. In particular, we encourage the empowerment of teachers during LA development, as this would strengthen the theoretical foundation of LA solutions and increase their interpretability and usability. Furthermore, we provide examples of how process data can be used to understand learning processes and generate more interpretable LA insights. Furthermore, we explore how LLMs could come into play in LA to generate interpretable insights, timely and actionable feedback, increase personalization, and support teachers’ tasks more broadly.

This project aims to create a virtual assistant with AI integration to improve Siemens Energy's internal processes. Using cloud-based technologies, microservice architecture, and large language models (LLMs), the project seeks to create a reliable, effective, and user-friendly assistant customized to Siemens Energy's requirements. The first significant business difficulty identified by the study was the time engineers had to spend looking for information in large volumes of company papers. The proposed virtual assistant responds with precision and context awareness to optimize productivity. The assistant uses a microservice architecture to guarantee scalability, flexibility, and integration for various use scenarios. Tasks like document retrieval, translation, summarization, and comparison can now be handled effectively. Utilizing Amazon Web Services (AWS) for cost-effectiveness and scalability, the backend is cloud-deployed, backed by a frontend created for natural user interaction. To increase precision and relevance, the system uses cutting-edge AI, such as vector databases and Retrieval Augmented Generation (RAG). The assistant expedites document management procedures, improves data accessibility, and reduces search time. The results highlight how it may enhance workflow efficiency for Siemens Energy engineers and how flexible it can be for future AI-driven applications.

This paper proposes the utilization of large language models as recommendations systems for museums. Since the aforementioned models lack the notion of context, they can’t work with temporal information that is often present in recommendations for cultural environments (e.g. special exhibitions or events). In this respect, the current work aims at enhancing the capabilities of large language models through a fine-tuning process that incorporates contextual information and user instructions. The resulting models are expected to be capable of providing personalized recommendations, aligned with user preferences and desires. More specifically, Generative Pre-trained Transformer 4, a knowledge-based large language model is fine-tuned and turned into a context-ware recommendation system, adapting its suggestions based on user input and specific contextual factors such as location, time of visit, and other relevant parameters. The effectiveness of the proposed approach is evaluated through certain user studies, which ensure an improved user experience and engagement within the museum environment.

Cognitive Warfare (CogWar) is another form of warfare where the goal is to exploit cognition facets to disrupt, undermine, influence or modify human decisions. Famous examples include “Operation Gidlock” meddling with the US 2016 election, and CogWar campaigns assisting the annexation of Crimea. As Large Language Model(LLM)-based applications expand, reaching more users and their cognitive states, stealth Malicious Cognitive Behavioral Tactics (MCBT) can be embedded in trusted application sessions to systematically profile, then alter, each individual’s cognition over time. Successful MCBT can create more cases like Snowden, mass shootings, or January6. This short-form paper discusses a novel MCBT threat model, a novel kill chain, and an on-going prototype for demonstration.

Although large language models demonstrate impressive capabilities, they sometimes generate irrelevant or nonsensical text, or produce outputs that deviate from the provided source input—an occurrence commonly referred to as hallucination. To mitigate this issue, we introduce a novel decoding method that incorporates both factual and hallucination prompts(DFHP), utilizing a contrastive output distribution to highlight the disparity in output probabilities between model predictions influenced by factual prompts and those affected by hallucination prompts. Experiments on both multiple-choice and text generation tasks show that our approach significantly enhances the factual accuracy of large language models without requiring additional training.

Sentiment analysis is a technique used to understand the publics’ opinion towards an event, product, or organization. For example, positive or negative opinion or attitude towards electric vehicle (EV) brands. This provides companies with valuable insight about the public's opinion of their products and brands. In the field of natural language processing (NLP), transformer models have shown great performances over the traditional machine learning algorithms. However, these models have not been explored extensively in the EV domain. EV companies are becoming signif-icant competitors in the automotive industry and are projected to cover up to 30% of the United States light vehicle market by 2030 [1]. In this study, we present a comparative study of large language models (LLMs) including bidirectional encoder representations from transformers (BERT), robustly optimized BERT approach and generalized autoregressive pretraining for language un-derstanding using Lucid motors and Tesla motors YouTube datasets. Results evidenced LLMs like BERT and her variants are off-the-shelf algorithms for sentiment analysis, specifically, when fi-ne-tuned. Furthermore, our findings presents the need for domain adaptation whilst utilizing LLMs. Finally, the experimental results showed that RoBERTa achieved consistent performance across the EV datasets with a F1 score of at least 92%.

The pursuit of Artificial Intelligence (AI) that emulates human cognitive processes is a cornerstone of ethical AI development, ensuring that emerging technologies can integrate seamlessly into societal frameworks requiring nuanced understanding and decision-making. Zero-Shot Instance Retrieval (ZSIR) stands at the forefront of this endeavour, potentially providing a robust platform for AI systems, particularly large visual language models, to demonstrate and refine cognition-aligned learning without the need for direct experience. In this paper, we critically evaluate current cognition alignment methodologies within traditional zero-shot learning paradigms, using visual attributes and word embedding generated by large AI models. We propose a unified similarity function that quantifies the cognitive alignment level, bridging the gap between AI processes and human-like understanding. Through extensive experimentation, our findings illustrate that this similarity function can effectively mirror the visual-semantic gap, steering the model towards enhanced performance in zero-shot instance retrieval. This work not only benchmarks the cognition alignment of AI, but also sets a new precedent for the development of visual language models attuned to the complexities of human cognition.}

This paper examines the potential of large language models, specifically ChatGPT, BingChat, and Bard, in enhancing the work of teachers in various educational settings. These models, powered by cutting-edge artificial intelligence, offer a wide range of applications that can aid educators in curriculum development, personalized instruction, student engagement, and administrative tasks. By harnessing the capabilities of these language models, teachers can not only streamline their workload but also improve the overall quality of education. This paper explores the ways in which these models can empower educators and revolutionize teaching practices.

Large Language Models (LLMs) have demonstrated promising capabilities to solve complex tasks in critical sectors such as healthcare. However, LLMs are limited by their training data which is often outdated, the tendency to generate inaccurate ("hallucinated") content and a lack of transparency in the content they generate. To address these limitations, retrieval augmented generation (RAG) grounds the responses of LLMs by exposing them to external knowledge sources. However, in the healthcare domain there is currently a lack of systematic understanding of which datasets, RAG methodologies and evaluation frameworks are available. This review aims to bridge this gap by assessing RAG-based approaches employed by LLMs in healthcare, focusing on the different steps of retrieval, augmentation and generation. Additionally, we identify the limitations, strengths and gaps in the existing literature. Our synthesis shows that proprietary models such as GPT-3.5/4 are the most used for RAG applications in healthcare. Also, there is a lack of standardized evaluation frameworks for RAG-based applications. In addition, the majority of the studies do not assess or address ethical considerations related to RAG in healthcare. It is important to account for ethical challenges that are inherent when AI systems are implemented in the clinical setting. Lastly, we highlight the need for further research and development to ensure responsible and effective adoption of RAG in the medical domain.

With the rapid growth of large language models (LLMs), there is increasing demand for memory and computation in LLMs. Recent efforts on post-training pruning of LLMs aim to reduce the model size and computation requirements, yet the performance is still sub-optimal. In this paper, we present a plug-and-play solution for post-training pruning of LLMs. The proposed solution has two innovative components: 1) Relative Importance and Activations (RIA), a new pruning metric that jointly considers the weight and activations efficiently on LLMs; and 2) Channel Permutation, a new approach to maximally preserve important weights under N:M sparsity. The two proposed components can be readily combined to further enhance the N:M semi-structured pruning of LLMs. Our empirical experiments show that RIA alone can already surpass all existing post-training pruning methods on prevalent LLMs, e.g., LLaMA ranging from 7B to 65B. Furthermore, N:M semi-structured pruning with channel permutation can even outperform the original LLaMA2-70B on zero-shot tasks, together with practical speed-up on specific hardware. Our code is available at: https://github.com/biomedical-cybernetics/Relative-importance-and-activation-pruning

Structured science summaries or research contributions using properties or dimensions beyond traditional keywords enhances science findability. Current methods, such as those used by the Open Research Knowledge Graph (ORKG), involve manually curating properties to describe research papers' contributions in a structured manner, but this is labor-intensive and inconsistent between the domain expert human curators. We propose using Large Language Models (LLMs) to automatically suggest these properties. However, it's essential to assess the readiness of LLMs like GPT-3.5, Llama 2, and Mistral for this task before application. Our study performs a comprehensive comparative analysis between ORKG's manually curated properties and those generated by the aforementioned state-of-the-art LLMs. We evaluate LLM performance through four unique perspectives: semantic alignment and deviation with ORKG properties, fine-grained properties mapping accuracy, SciNCL embeddings-based cosine similarity, and expert surveys comparing manual annotations with LLM outputs. These evaluations occur within a multidisciplinary science setting. Overall, LLMs show potential as recommendation systems for structuring science, but further finetuning is recommended to improve their alignment with scientific tasks and mimicry of human expertise.

Obsessive-Compulsive Disorder (OCD) is a mental health condition marked by recurrent intrusive thoughts or sensations that compel individuals to perform repetitive behaviors or mental acts. Obsessions and compulsions significantly disrupt daily life and cause considerable distress. Early identification and intervention improve long-term outcomes. This study aimed to evaluate the ability of four advanced artificial intelligence models (ChatGPT-3.5, ChatGPT-4, Claude, and Bard) to accurately recognize OCD compared to human professionals and to assess recommended therapies and stigma attributions. This study was conducted during March 2024 utilizing 12 vi-gnettes. Each vignette depicted a client, either a young adult or a middle-aged male or female, attending an initial therapy session. Each vignette was evaluated ten times, resulting in 480 evaluations. The results were compared with those of a human sample of 514 psychotherapists, as reported by Canavan. Significant differences were found. AI models demonstrated higher OCD recognition rates and confidence levels than human professionals and showed 100% confidence in recognition, compared to 87% among psychotherapists. AI models also recommended evi-dence-based interventions more frequently, with ChatGPT-3.5 and Claude at 100%, ChatGPT-4 at 90%, and Bard at 60%, compared to 61.9% among psychotherapists. Additionally, AI models ex-hibited significantly lower stigma and danger estimations, though both AI and psychotherapists demonstrated high willingness to treat the described cases. The findings suggest that AI models surpass human professionals in recognizing OCD and recommending evidence-based treatments while also demonstrating lower stigma. These results highlight the potential of AI tools to enhance OCD diagnosis and treatment in clinical settings.

The rise of abusive language on social media is a significant threat to mental health and 1 social cohesion. For Bengali speakers, the need for effective detection is critical. However, current 2 methods fall short in addressing the massive volume of content. Improved techniques are urgently 3 needed to combat online hate speech in Bengali. Traditional machine learning techniques, while 4 useful, often require large, linguistically diverse datasets to train models effectively. This paper 5 addresses the urgent need for improved hate speech detection methods in Bengali, aiming to fill the 6 existing research gap. Contextual understanding is crucial in differentiating between harmful speech 7 and benign expressions. Large language models (LLMs) have shown state-of-the-art performance in 8 various natural language tasks due to their extensive training on vast amounts of data. We explore the 9 application of LLMs, specifically GPT-3.5 Turbo and Gemini 1.5 Pro, for Bengali hate speech detection 10 using Zero-Shot and Few-Shot Learning approaches. Unlike conventional methods, Zero-Shot 11 Learning identifies hate speech without task-specific training data, making it highly adaptable to new 12 datasets and languages. Few-Shot Learning, on the other hand, requires minimal labeled examples, 13 allowing for efficient model training with limited resources. Our experimental results show that 14 LLMs outperform traditional approaches. In this study, we evaluated GPT-3.5 Turbo and Gemini 1.5 15 Pro on multiple datasets. To further enhance our study, we considered the distribution of comments 16 in different datasets and the challenge of class imbalance, which can affect model performance. The 17 BD-SHS dataset consists of 35,197 comments in the training set, 7,542 in the validation set, and 7,542 18 in the test set. The Bengali Hate Speech Dataset v1.0 & v2.0 includes comments distributed across 19 various hate categories: personal hate (629), political hate (1,771), religious hate (502), geopolitical hate 20 (1,179), and gender abusive hate (316). The Bengali Hate Dataset comprises 7,500 non-hate and 7,500 21 hate comments. GPT-3.5 Turbo achieved impressive results with 97.33%, 98.42%, and 98.53% accuracy. 22 In contrast, Gemini 1.5 Pro showed lower performance across all datasets. Specifically, GPT-3.5 Turbo 23 excelled with significantly higher accuracy compared to Gemini 1.5 Pro. These outcomes highlight a 24 6.28% increase in accuracy compared to traditional methods, which achieved 92.25%. Our research 25 contributes to the growing body of literature on LLM applications in natural language processing, 26 particularly in the context of low-resource languages.

Recommender Systems (RecSys) are crucial in managing information overload and enhancing user satisfaction across various digital platforms, including e-commerce and entertainment. Evolving from traditional models to Deep Neural Networks (DNNs) and, more recently, Large Language Models (LLMs), these systems leverage sophisticated algorithms to analyze user behaviors and preferences. LLMs, such as GPT-4, are trained on extensive datasets to comprehend and generate natural language, significantly advancing their ability to deliver personalized recommendations. This tutorial explores the transformative impact of LLMs on RecSys, discussing their development, application in handling complex datasets, and the integration of contextual insights. Real-world examples illustrate how LLMs enhance recommendation accuracy and user experience, highlighting challenges and future directions in the field.

In this paper, we propose "Synergized Efficiency Optimization for Large Language Models" (SEO-LLM), a groundbreaking approach that integrates advanced data utilization and model compression techniques to significantly enhance the performance, efficiency, and scalability of large language models (LLMs). Our method synergistically combines Adaptive Data Augmentation (ADA), Transfer- Active Learning (TAL), Adaptive Iterative Pruning (AIP), and Synergistic Quantization and Distillation (SQD). These components work together to reduce the training data requirement by 30%, compress model size by 67.6%, and improve inference speed by up to 50%, while preserving or even enhancing model accuracy across various NLP tasks. ADA dynamically adjusts augmentation strategies to optimize model generalization, while TAL leverages pre-trained models to focus learning on the most informative data samples. AIP intelligently prunes less significant weights, and SQD harmonizes quantization with knowledge distillation to achieve high compression rates without significant performance loss. The synergy between these techniques makes SEO-LLM a robust solution for deploying LLMs in resource-constrained environments, maintaining state-of-the-art performance with a fraction of the computational and data resources.

Large Language Models (LLMs) like GPT-4 and mBERT have revolutionized natural languageprocessing (NLP) by providing multilingual capabilities, making it possible to develop models thathandle diverse linguistic inputs across various languages. However, despite these advances, thereremains a noticeable performance gap between how well these models perform in high-resourcelanguages such as English and low-resource languages such as Nepali or Malagasy. We term thisphenomenon the "Babel Effect," highlighting the disproportionate performance that arises fromdifferences in resource availability across languages.This paper aims to explore the root causes of these performance discrepancies in LLMs, focusingon the underlying challenges in tokenization, training, and data scarcity. We utilize cross-lingualbenchmarks, such as XGLUE and TyDiQA, to quantify these performance variations and examinethem in detail. Furthermore, we propose solutions, including enhancing tokenization strategies,employing data augmentation techniques, and refining fine-tuning methods. The paper concludeswith a discussion on how these improvements can mitigate the Babel Effect and lead to more equitablelanguage modeling across diverse linguistic contexts.

The growing dependence on large language models (LLM)s highlights the urgent need to deepen trust in these technologies. Regular, rigorous validation of their expertise, especially in nuanced and intricate scenarios, is essential to ensure their readiness for clinical applications. Our study pioneers the exploration of LLM utility in the field of cardiology. We stand at the cusp of a transformative era where mature AI and LLMs, notably ChatGPT, GPT-4, and Google Bard, are poised to influence healthcare significantly. Recently, we put three available LLMs, OpenAI's ChatGPT-3.5, GPT-4.0, and Google's Bard, to the test against a significant Polish medical special-ization licensing exam (PES). The exams cover the scope of completed specialist training, focusing on diagnostic and therapeutic procedures, excluding invasive medical procedures and interven-tions. In our analysis, GPT-4 consistently outperformed the others, ranking first, with and Google Bard and ChatGPT- 3.5 following, respectively. The performance metrics underscore GPT-4's no-table potential in medical applications. Given a score improvement of over 23.5 % between two AI models released just four months apart, clinicians must stay informed and up-to-date about these rapidly evolving tools and their potential applications to clinical practice. Our results provide a snapshot of the current capabilities of these models, highlighting the nuanced performance dif-ferences when confronted with identical questions

This study presents a novel approach to enhance Large Language Models (LLMs) like Alpaca by dynamically integrating real-time information. This method addresses the issue of content hallucination and data relevancy by automatically collecting and integrating current data from credible sources into model prompts. Experiments show a significant improvement in accuracy and a decrease in content hallucination, with a manageable increase in response time. The research underscores the potential of real-time data integration in making LLMs more accurate and contextually relevant, setting a foundation for future advancements in dynamic data processing in AI.

Generating executable code from natural language instructions using Large Language Models (LLMs) poses challenges such as semantic ambiguity and understanding task- specific contexts. To address these issues, we propose a system called DemoCraft, which enhances code generation by leveraging in-context learning and demonstration selection, combined with latent concept learning. Latent concept learning introduces additional concept tokens, which are trainable embeddings that capture task-specific knowledge. We then test our system on two major datasets: MBPP and HumanEval. Our experimental results demonstrate that the proposed system achieves an approximate 2x increase in the pass@k metric compared to baseline models. Furthermore, we introduce two novel evaluation metrics: correctness@k and similarity@k. Our empirical studies indicate that our system attains nearly a 3x improvement in these metrics as well.

Large language models (LLMs) are essential to the field of natural language processing, and as their applications expand, security risks have become increasingly prominent. This paper introduces a novel benchmark for evaluating LLMs security, termed CS-Eval, designed to effectively assess the models' ability to address vulnerabilities. CS-Eval targets seven key security risks: ethical dilemmas, marginal topics, error detection, detailed event handling, cognitive bias, logical reasoning, and privacy identification, and establishes a Multi-Security Hazard Dataset (MSHD). The evaluated models include GPT-4o, Llama-3-70B, Claude-3-Opus, ERNIE-4.0, Abab-6.5, Qwen1.5-110B, Gemini-1.5-Pro, Doubao-Pro, SenseChat-V5, and GLM-4. We analyzed each model's performance in relation to these security risks and provided recommendations for improvement. Experimental results demonstrate varying levels of effectiveness across models, with GPT-4o exhibiting the best overall performance. Moreover, the relationship between security enhancement and model capa-bility is nonlinear, indicating that improving safety requires a multifaceted approach, considering various factors in both development and application.

Tactics, Techniques, and Procedures (TTPs) constitute the most valuable aspect of Cyber Threat Intelligence (CTI). However, TTPs are often implicit in unstructured text, necessitating manual analysis by field experts. Automating the classification of TTPs from unstructured text is a crucial task in contemporary research. MITRE ATT&CK serves as the de facto standard for studying TTPs. Existing research constructs classification datasets based on its procedural examples for tactics and techniques. However, due to a significant proportion of small sample categories, a long-tail phenomenon exists, leading to a highly imbalanced sample distribution. Consequently, more research concentrates on categories with relatively abundant samples. This paper proposes a method that combines ChatGPT data augmentation with Instruction Supervised Fine-Tuning of open large language models. This approach offers a solution for TTPs classification in few-shot learning scenarios, achieving coverage of 625 technical categories. The Precision, Recall, and F1 scores reach 86.2%, 89.9%, and 87.3%, respectively.

Patient symptoms play a critical role in disease progression and diagnosis, yet they are most often captured in unstructured clinical notes. This study explores use of large language models (LLMs) for extracting patient symptoms from clinical notes, comparing their performance against rule-based information extraction (IE) systems, like BioMedICUS. By fine-tuning an LLM on diverse corpora from multiple healthcare institutions, we aimed to improve symptom extraction accuracy and efficiency with symptoms related to acute and post-acute sequelae of SARS-CoV-2. We also conducted prevalence analysis to highlight significant differences in symptom prevalence across corpora and performed fairness analysis to assess the model's equity across race and gender. Our findings indicate that while LLMs can match the effectiveness of rule-based IE methods, they face significant challenges related to demographic bias and generalizability due to variability in training corpora. This evaluation revealed overfitting and insufficient generalization, especially when models were trained predominantly on limited datasets with single annotator bias. This study also revealed that LLMs offer substantial advantages in efficiency, adaptability, and scalability for biomedical IE, marking a transformative shift in clinical data extraction processes. These results provide real-world evidence of the necessity for diverse, high-quality training datasets and robust annotation processes to enhance LLM’s performance and reliability in clinical applications.

The rapid increase in unstructured data across various fields has made multi-document comprehension and summarization a critical task. Traditional approaches often fail to capture relevant context, maintain logical consistency, and extract essential information from lengthy documents. This paper explores the use of Long-context Large Language Models (LLMs) for multi-document summarization, demonstrating their exceptional capacity to grasp extensive connections, provide cohesive summaries, and adapt to various industry domains and integration with enterprise applications/systems. The paper discusses the workflow of multi-document summarization for effectively deploying long-context LLMs, supported by case studies in legal applications, enterprise functions such as HR, finance, and sourcing, as well as in the medical and news domains. These case studies show notable enhancements in both efficiency and accuracy. Technical obstacles, such as dataset diversity, model scalability, and ethical considerations like bias mitigation and factual accuracy, are carefully analyzed. Prospective research avenues are suggested to augment the functionalities and applications of long-context LLMs, establishing them as pivotal tools for transforming information processing across diverse sectors and enterprise applications.

Large language models (LLMs) have shown impressive performance and versatility. However, their billions of parameters and high computational costs hinder the development of personalized and privacy-preserving AI assistants operating locally on user devices. In this work, we explored the potential of pruning LLMs to create lightweight models suitable for user devices, using the moderate-sized Llama-2 7B model as an example. By adopting a simple yet effective pruning method, we found that up to 60% of the weights in the Llama-2 7B model could be pruned without significantly impairing its language modeling capabilities. Furthermore, despite occasional factual inaccuracies, the pruned model at the sparsity limit generated fluent and helpful answers to daily queries, demonstrating its feasibility of on-device AI assistants. These inaccuracies might originate from forgetting or hallucination due to pruning. We proposed a simple protocol to distinguish between the two mechanisms, as well as future directions to improve the pruned models for local AI assistants.

The accurate diagnosis and effective treatment of mental health disorders such as depression remain challenging owing to the complex underlying causes and varied symptomatology. Traditional information extraction methods struggle to adapt to evolving diagnostic criteria such as the Diagnostic and Statistical Manual of Mental Disorders fifth edition (DSM-5) and to contextualize rich patient data effectively. This study proposes a novel approach for enhancing information extraction from mental health data by integrating medical knowledge graphs and large language models (LLMs). Our method leverages the structured organization of knowledge graphs specifically designed for the rich domain of mental health combined with the powerful predictive capabilities and zero-shot learning abilities of LLMs. This research enhances the quality of knowledge graphs through entity linking and demonstrates superiority over traditional information extraction techniques, making a significant contribution to the field of mental health. It enables a more fine-grained analysis of the data and the development of new applications. Our approach redefines the manner in which mental health data are extracted and utilized. By integrating these insights with existing healthcare applications, the groundwork is laid for the development of real-time patient monitoring systems. The performance evaluation of this knowledge graph highlights its effectiveness and reliability, indicating significant advancements in automating medical data processing and depression management.

(1) Background: Artificial Intelligence (AI) models have shown potential in various educational contexts. However, their utility in explaining complex biological phenomena, such as Intrinsically Disordered Proteins (IDPs), requires further exploration. This study empirically evaluated the performance of various Large Language Models (LLMs) in the educational domain of IDPs. (2) Methods: Four LLMs, GPT-3.5, GPT-4, GPT-4 with Browsing, and Google Bard (PaLM 2), were assessed using a set of IDP-related questions. An expert evaluated their responses across five categories: accuracy, relevance, depth of understanding, clarity, and overall quality. Descriptive statistics, ANOVA, and Tukey's honesty significant difference tests were utilized for analysis. (3) Results: The GPT-4 model consistently outperformed the others across all evaluation categories. Although GPT-4 and GPT-3.5 were not statistically significantly different in performance (p>0.05), GPT-4 was preferred as the best response in 13 out of 15 instances. The AI models with browsing capabilities, GPT-4 with Browsing and Google Bard (PaLM 2) displayed lower performance metrics across the board with statistically significant differences (p<0.0001). (4) Conclusion: Our findings underscore the potential of AI models, particularly LLMs such as GPT-4, in enhancing scientific education, especially in complex domains such as IDPs. Continued innovation and collaboration among AI developers, educators, and researchers are essential to fully harness the potential of AI for enriching scientific education.

Virtual assistants (VAs) have gained widespread popularity across a wide range of applications, and the integration of Large Language Models (LLMs) such as ChatGPT has opened up new possibilities for developing even more sophisticated VAs. However, this integration poses new ethical issues and challenges that must be carefully considered, particularly as these systems are increasingly used in public services: transfer of personal data, decision-making transparency, potential biases, and privacy risks. This paper, an extension of the work presented at IberSPEECH 2022, analyzes the current regulatory framework for AI-based VAs in Europe and delves into ethical issues in depth, examining potential benefits and drawbacks of integrating LLMs with VAs. Based on the analysis, this paper argues that the development and use of VAs powered by LLMs should be guided by a set of ethical principles that prioritize transparency, fairness, and harm prevention. The paper presents specific guidelines for the ethical use and development of this technology, including recommendations for data privacy, bias mitigation, and user control. By implementing these guidelines, the potential benefits of VAs powered by LLMs can be fully realized while minimizing the risks of harm and ensuring that ethical considerations are at the forefront of the development process.

ChatGPT, a Large Language Model (LLM) utilizing Natural Language Processing (NLP), has caused concerns about its impact on job sectors, including cybersecurity. In this study, we assessed ChatGPT’s impacts in non-managerial cybersecurity roles using the NICE Framework and Technological Displacement theory. We also explored its potential in passing top cybersecurity certification exams. Findings revealed ChatGPT’s promise in streamlining some jobs, especially those requiring memorization. Moreover, we highlighted ChatGPT’s challenges and limitations, such as ethical implications, LLM limitations, and Artificial Intelligence (AI) security. The study suggests that LLMs like ChatGPT could transform the cybersecurity landscape, causing job losses, skill obsolescence, labor market shifts, and mixed socioeconomic impacts. We recommend a shift in focus from memorization to critical thinking, and collaboration between LLM developers and cybersecurity professionals.

The increased complexity of modern systems is calling for an integrated and comprehensive approach to system design and development and in particular, a shift towards Model-Based Systems Engineering (MBSE) approaches for system design. The requirements that serve as the foundation for these intricate systems are still primarily expressed in Natural Language (NL), which can contain ambiguities and inconsistencies that hinder their direct translation into models. The colossal developments in the field of Natural Language Processing (NLP) in general and Large Language Models (LLMs) in particular can serve as an enabler for the conversion of NL requirements into semi-machine-readable requirements. This is expected to facilitate their standardization and use in a model-based environment. This paper discusses a two-fold strategy for converting NL requirements into semi-machine-readable requirements using language models. The first approach involves creating a requirements table by extracting information from free-form NL requirements. The second approach is an agile methodology that facilitates the identification of boilerplate templates for different types of requirements based on observed linguistic patterns. For this study, three different LLMs were utilized. Two of these models were fine-tuned versions of Bidirectional Encoder Representations from Transformers (BERT), specifically aeroBERT-NER and aeroBERT-Classifier, which were trained on annotated aerospace corpora. Another LLM, called flair/chunk-english, was utilized to identify sentence chunks present in NL requirements. All three language models were utilized together to achieve the standardization of requirements. To demonstrate the effectiveness of the methodologies, requirements from Parts 23 and 25 of Title 14 Code of Federal Regulations (CFRs) were employed, and a total of two, five, and three boilerplate templates were identified for design, functional, and performance requirements, respectively.

: Multimodal Large Language Models (MLLMs) harness comprehensive knowledge spanning text, images, and audio to adeptly tackle complex problems, including zero-shot in-context learning scenarios. This study explores the ability of MLLMs in visually solving the Traveling Salesman Problem (TSP) and Multiple Traveling Salesman Problem (mTSP) using images that portray point distributions on a two-dimensional plane. We introduce a novel approach employing multiple specialized agents within the MLLM framework, each dedicated to optimizing solutions for these combinatorial challenges. Our experimental investigation includes rigorous evaluations across zero-shot settings and introduces innovative multi-agent zero-shot in-context scenarios. The re-sults demonstrated that both multi-agent models—Multi-Agent 1, which includes the Initializer, Critic, and Scorer agents, and Multi-Agent 2, which comprises only the Initializer and Critic agents—significantly improved solution quality for TSP and mTSP problems. Multi-Agent 1 ex-celled in environments requiring detailed route refinement and evaluation, providing a robust framework for sophisticated optimizations. In contrast, Multi-Agent 2, focusing on iterative re-finements by the Initializer and Critic, proved effective for rapid decision-making scenarios. These experiments yield promising outcomes, showcasing the robust visual reasoning capabilities of MLLMs in addressing diverse combinatorial problems. The findings underscore the potential of MLLMs as powerful tools in computational optimization, offering insights that could inspire further advancements in this promising field.

The proliferation of wearable technology enables the generation of vast amounts of sensor data, offering significant opportunities for advancements in health monitoring, activity recognition, and personalized medicine. However, the complexity and volume of this data present substantial challenges in data modeling and analysis, which have been tamed with approaches spanning time series modeling to deep learning techniques. The latest frontier in this domain is the adoption of Large Language Models (LLMs), such as GPT-4 and Llama, for data analysis, modeling, understanding, and generation of human behavior through the lens of wearable sensor data. This survey explores current trends and challenges in applying LLMs for sensor-based human activity recognition and behavior modeling. We discuss the nature of wearable sensors data, the capabilities and limitations of LLMs to model them and their integration with traditional machine learning techniques. We also identify key challenges, including data quality, computational requirements, interpretability, and privacy concerns. By examining case studies and successful applications, we highlight the potential of LLMs in enhancing the analysis and interpretation of wearable sensors data. Finally, we propose future directions for research, emphasizing the need for improved preprocessing techniques, more efficient and scalable models, and interdisciplinary collaboration. This survey aims to provide a comprehensive overview of the intersection between wearable sensors data and LLMs, offering insights into the current state and future prospects of this emerging field.

Estimating Visual Quality of Street Space (VQoSS) is pivotal for urban design, environmental sustainability, civic engagement, etc. Recent advancements, notably in deep learning, have enabled large-scale analysis. However, traditional deep learning approaches are hampered by extensive data annotation requirements and limited adaptability across diverse VQoSS tasks. Multimodal Large Language Models (MLLMs) have recently demonstrated proficiency in various computer vision tasks, positioning them as promising tools for automated VQoSS assessment. In this paper, we pioneer the application of MLLMs to VQoSS change estimation, with our empirical findings affirming their effectiveness. In addition, we introduce Street Quality GPT (SQ-GPT), a model that distills knowledge from the current most powerful but inaccessible (not free) GPT-4, requiring no human efforts. SQ-GPT approaches GPT-4’s performance and is viable for large-scale VQoSS change estimation. In a case study of Nanjin, we showcase the practicality of SQ-GPT and knowledge distillation pipeline. Our work promises to be a valuable asset for future urban studies research.

The development of intelligent medical service robots for patient care presents significant chal-lenges, particularly in integrating diverse knowledge sources and enabling robots to autono-mously perform tasks in dynamic and unpredictable healthcare environments. This study in-troduces a novel framework that combines large language models (LLMs) with healthcare-specific knowledge and robotic operations to enhance autonomous task execution for a Robotic Health Attendant (RHA). Utilizing OpenAI’s ChatGPT, the RHA processes structured information about patient care protocols and unstructured human inputs to generate context-aware robot actions. A prototype system was tested in a simulated patient room where the RHA successfully performed both simple individual actions and complex tasks involving the execution of multiple actions, based on real-time dialogues with the LLM and predefined task specifications. The results demonstrate the potential of LLMs to reduce the reliance on hardcoded logic and provide healthcare professionals with the ability to interact with robotic systems through natural lan-guage.

Large language models (LLMs) excel in providing natural language responses that sound authoritative, reflect knowledge of the context area, and can present from a range of varied perspectives. Agent Based Models and Simulation consist of simulated agents that interact within a simulated environment to explore societal, social, and ethical, among other, problems. Agents generate large volumes of data over time and discerning useful and relevant content is an onerous task. LLMs can help in communicating agents’ perspectives on key events by providing natural language narratives. However, these narratives need to be factual, transparent, and reproducible. To this end, we present a structured narrative prompt for sending queries to LLMs. Chi-square tests and Fisher’s Exact tests are applied to assess statistically significant difference in sentiment scores of the narrative messages between simulation generated narratives, ChatGPT-generated narratives, and real tweets. The narrative prompt structure effectively yields narratives with the desired components from ChatGPT. This structure is expected to be extensible across LLMs. In 14 out of 44 categories, ChatGPT generated narratives which has sentiment scores that were not discernibly different, in terms of statistical significance (alpha level of 0.05), from the sentiment expressed in real tweets. Three outcomes are provided: (1) a list of benefits and challenges for LLMs in narrative generation; (2) a structured prompt for requesting narratives of a LLM based on simulated agents’ information; and (3) an assessment of statistical significance in the sentiment prevalence of the generated narratives compared to real tweets. This indicates significant promise in the utilization of LLMs for helping to connect simulated agent’s experiences with real people.

The predominance of English and Latin-based large language models (LLMs) has led to a notable deficit in native Arabic LLMs. This discrepancy is accentuated by the prevalent inclusion of English tokens in existing Arabic models, detracting from their efficacy in processing native Arabic's intricate morphology and syntax. Consequently, there is a theoretical and practical imperative for developing LLMs predominantly focused on Arabic linguistic elements. To address this gap, this paper proposes ArabianGPT, a series of transformer-based models within the ArabianLLM suite designed explicitly for Arabic. These models, including ArabianGPT-0.1B and ArabianGPT-0.3B, vary in size and complexity, aligning with the nuanced linguistic characteristics of Arabic. The AraNizer tokenizer, integral to these models, addresses the unique morphological aspects of Arabic script, ensuring more accurate text processing. Empirical results from fine-tuning the models on tasks like sentiment analysis and summarization demonstrate significant improvements. For sentiment analysis, the fine-tuned ArabianGPT-0.1B model achieved a remarkable accuracy of 95%, a substantial increase from the base model's 56%. Similarly, in summarization tasks, fine-tuned models showed enhanced F1 scores, indicating improved precision and recall in generating concise summaries. Comparative analysis of fine-tuned ArabianGPT models against their base versions across various benchmarks reveals nuanced differences in performance, with fine-tuning positively impacting specific tasks like question answering and summarization. These findings underscore the efficacy of fine-tuning in aligning ArabianGPT models more closely with specific NLP tasks, highlighting the potential of tailored transformer architectures in advancing Arabic NLP.

Good communication skills are essential for prospective workers, particularly college students, to clearly convey their abilities and understanding during job interviews. Despite the high demand for strong communication skills among employers, many students struggle with inadequate experience and skills, leading to reduced confidence and lower chances of securing desired jobs. Existing tools and resources for improving communication skills are limited, especially free ones, leaving students underprepared and less confident in interviews. To address this issue, a comprehensive solution is needed that provides free, customizable tools to enhance students' interview communication skills. Invisor is a web-based self-interview training application that leverages AI and Machine Learning to analyze users' behavior, responses, and facial expressions during simulated interviews. Accessible via desktop, laptop, or mobile devices, Invisor offers interactive and voice-recorded feedback to help students prepare for interviews effectively. By employing AI technologies, Invisor generates personalized interview questions based on job descriptions, enhancing the realism of the interview experience and providing tailored feedback to improve users' responses. Developed using agile methodologies, Invisor has undergone iterative enhancements, integrating user and stakeholder feedback promptly to improve usability and functionality. The combination of AI-driven simulations and agile development practices enables Invisor to meet current user needs and anticipate future demands, providing a robust tool for interview preparation in a competitive job market. This paper outlines the development and implementation of Invisor, demonstrating its effectiveness in enhancing interview skills and confidence among students and job seekers.

This research aims to explore the integration method of deep learning and large language models in speech recognition to improve the system’s recognition accuracy and ability to handle complex contexts. Deep neural network (DNN), convolutional neural network (CNN), long short-term memory network (LSTM) and Transformer-based large language model are used to build an integrated acoustic and language model framework. Experiments on TIMIT, LibriSpeech and Common Voice datasets show that the ensemble model shows significant improvements in both word error rate (WER) and real-time factor (RTF) compared to traditional models. Especially in terms of adaptability to multiple languages and accent changes, the model shows superior performance. The conclusion shows that through technology integration, the performance of the speech recognition system in complex environments can be effectively improved, providing a new direction for the future development of speech recognition technology.

In February of 2024, Dr. Jun Zhuang and Dr. Casey Kennington published a paper (Zhuang and Kennington, 2024) in which they classified large language model (LLM) survey papers into different taxonomies utilizing graph learning. In this paper, I evaluate the dataset they created and used and propose that in its current state, there is not enough samples to classify other survey papers into specific categories.

This research investigates the potential application of generative language models, especially ChatGPT, in aviation safety analysis as a means to enhance the efficiency of safety analyses and accelerate the time it takes to process incident reports. In particular, ChatGPT was leveraged to generate incident synopses from narratives, which were subsequently compared with ground truth synopses from the Aviation Safety Reporting System (ASRS) dataset. The comparison was facilitated by using embeddings from Large Language Models (LLMs), with aeroBERT demonstrating the highest similarity due to its aerospace-specific fine-tuning. A positive correlation was observed between synopsis length and their cosine similarity. In a subsequent phase, human factor issues involved in incidents as identified by ChatGPT were compared to human factor issues identified by safety analysts. A concurrence rate of 61% was found, with ChatGPT demonstrating a cautious approach towards attributing human factor issues. Finally, the model was used to attribute incidents to relevant parties. As no dedicated ground truth column existed for this task, a manual evaluation was conducted. ChatGPT attributed the majority of incidents to the Flight Crew, ATC, Ground Personnel, and Maintenance. This study opens new avenues for leveraging AI in aviation safety analysis.

In a trip planning service, travelers first set a starting point, a destination, and a sequential list of specific points of interest types (e.g., museums, restaurants, and parks). Based on this information, the service searches the spatial database to customize the best travel itinerary for the tourist. However, in previous studies, planners only considered the time factor when designing the optimal route, and failed to fully consider the quality of each point of interest. In this study, we specifically leveraged the capabilities of large language model to parse and respond to complex travel-related user queries. To apply large language model to route planning, we fine-tuned the model to understand geotagging and user travel preferences. We have introduced a novel graph search algorithm combined with large language model output, which optimizes the route search process to provide optimal travel recommendations by taking into account various factors such as distance length, budget constraints, and popularity of tourist attractions. In addition, we have integrated real-time traffic data and historical travel data to further improve the prediction accuracy and application usefulness of the model. In the experimental validation phase, we designed a series of benchmarks to compare the performance of the system with traditional algorithms and other machine learning-based route planning methods. Experimental results show that our model has a significant improvement compared with traditional methods in improving the speed and accuracy of path selection.

Despite the growing capabilities of large language models, concerns exist about the biases they develop. In this paper, we propose a novel, automated mechanism for debiasing through specified dataset augmentation in the lens of bias producers that can be useful in a variety of industries, especially ones that are “restricted” and have limited data. We consider that bias can occur due to intrinsic model architecture and dataset quality. The two aspects are evaluated using two different metrics we created. We show that our dataset augmentation algorithm reduces bias as measured by our metrics.

This study evaluates the ability of six popular chatbots; ChatGPT-3.5, ChatGPT-4.0, Gemini, Copilot, Chatsonic, and Perplexity to provide reliable answers to questions concerning keratoconus. Methods: Chatbots responses were assessed using mDISCERN and Global Quality Score (GQS) metrics. Readability was evaluated using nine validated readability assessments. We also addressed the quality and accountability of websites from which the questions originated. Results: We analyzed 20 websites, 65% "Private practice or independent user" and 35% "Official patient education materials." The mean JAMA Benchmark score was 1.5±0.68, indicating low accountability. Reliability, measured using mDISCERN, ranged from 42.9±3.16 (ChatGPT-3.5) to 46.95±3.53 (Copilot). The most frequent question was "What is Keratoconus?" with 70% of websites providing relevant information. This received the highest mDISCERN score (49.33±4.96) and a relatively high GQS score (3.50±0.55), with an Automated Readability Level Calculator score of 13.17±2.13. Moderate positive correlations were determined between the website numbers and both mDISCERN (?=0.265,?=0.25) and GQS (?=0.453,?=0.05) scores. The quality of information, assessed using the GQS, ranged from 3.01±0.51 (ChatGPT-3.5) to 3.3±0.65 (Gemini) (p=0.34). The differences between the texts were statistically significant. Gemini emerged as the easiest to read, while ChatGPT-3.5 and Perplexity were the most difficult. Based on mDISCERN scores, Gemini and Copilot exhibited the highest percentage of responses in the "Good" range (51-62 points). For the GQS, the Gemini model exhibited the highest percentage of responses in the "Good" quality range, with 40% of its responses scoring 4-5. Conclusions: While all chatbots performed well, Gemini and Copilot showed better reliability and quality. However, their readability often exceeded recommended levels. Continuous improvements are essential to match information with patients' health literacy for effective use in ophthalmology.

This paper explores the role of open-source large language models in IoT cybersecurity world. The threats of malicious activity on the Internet and the loss of private information are very real and lead to serious consequences. The purpose of this paper is to investigate how open source-large language models can help to defend against the growing threat of cyber-crimes. We conducted our experiments in two directions. The first one is a security assistant that helps with cybersecurity best practices advices. The second one is a how large language model can simulate a vulnerable IoT system. For both types of experiments, the interactive mode of operation of the language model is used. In the context of the cybersecurity research, a major advantage of the locally installed open-sourced large language models is that they do not share sensitive data with a remote system in a cloud. The paper concludes by discussing the potential impact of open-source large language models on cybersecurity research and recommends future research directions.

According to a recent study by OpenAI, Open Research, and the University of Pennsylvania, large language models (LLMs) based on artificial intelligence (AI), such as Generative Pretrained Transformers (GPTs), may have potential implications on the job market, specifically regarding job occupations that demand writing or programming skills. The research points out Interpreters and Translators as one of the main job occupations with greater exposure to AI in the US job market (76.5%). This article, following a mixed-methods survey-based research, provides insights into the Portuguese Language Service Providers (LSPs) awareness and knowledge about AI, specifically regarding neural machine translation (NMT) and large language models (LLM), its actual use and usefulness, as well as their potential influence on work performance and labour market. The results show that most professionals are unable to identify if AI and/or automation technologies support the tools that are most used in the profession. The usefulness of AI is essentially low to moderate and the professionals who are less familiar with it and less knowledgeable also demonstrate lack of trust in it. Two thirds of the sample estimate negative or very negative effects of AI in the profession, expressing as main concerns the devaluation and replacement of experts, the reduction of income, and the reconfiguration of the career of translator to mere post-editors.

Software systems are often maintained by a group of experienced software developers, in order to ensure that faults which may bring the system down are less likely. Large turnover in organizations such as CERN makes it important to think of ways of onboarding newcomers on a technical project rapidly. This paper focuses on optimizing the way that people get up-to-speed on the business logic and technologies used on the project, using a knowledge-imbued large language model that is enhanced using domain-specific knowledge from the group or team’s internal documentation.

Since the beginning of the internet era, there has been an explosion of growth in structured data (such as numbers, symbols, and labels) as well as unstructured data (including images, videos, and text). Efficient and accurate mixed query of these two types of data is a key technology to achieve high-quality information retrieval, and it is also a major challenge that needs to be solved urgently in the industry. In this study, we employ an advanced Transformer model that combines strategies and fine-tuning techniques for multi-task learning. Specifically, the model is first pre-trained on a large-scale, general-purpose dataset to learn different types of data representations and basic language comprehension skills. After that, we fine-tuned the parameters of the model to better suit these specific data processing tasks for specific application scenarios, such as image annotation, video content analysis, and structured data query. At the heart of the model is the self-attention mechanism, which allows the model to automatically emphasize the important parts and ignore irrelevant information when processing the input data. In addition, we have introduced task-specific adaptation layers that are designed to add additional processing power to the original Transformer architecture, such as a semantic analysis layer for unstructured text data and a relational extraction layer for structured data. This combination of general pre-training and task-specific fine-tuning allows the model to flexibly process and integrate information from different data sources, improving processing efficiency and accuracy. Experimental results show that the model performs well in a variety of data processing tasks, significantly improves the accuracy and efficiency of information retrieval, and verifies the strong potential and adaptability of large language models in processing mixed data types.

As large language models (LLMs) continue to advance, evaluating their comprehensive capabilities becomes significant for their application in various fields. This research study comprehensively evaluates the language, vision, speech, and multimodal capabilities of GPT-4o. The study employs standardized exam questions, reasoning tasks, and translation assessments to assess the model's language capability. Additionally, GPT-4o's vision and speech capabilities are tested through image classification and object recognition tasks, as well as accent classification. The multimodal evaluation assesses the model's performance in integrating visual and linguistic data. Our findings reveal that GPT-4o demonstrates high accuracy and efficiency across multiple domains in language and reasoning capabilities, excelling in tasks that require few-shot learning. GPT-4o also provides notable improvements in multimodal tasks compared to its predecessors. However, the model shows variability and faces limitations in handling complex and ambiguous inputs, particularly in audio and vision capabilities. This paper highlights the need for more comprehensive benchmarks and robust evaluation frameworks, encompassing qualitative assessments involving human judgment as well as error analysis. Future work should focus on expanding datasets, investigating prompt-based assessment, and enhancing few-shot learning techniques to test the model's practical applicability and performance in real-world scenarios.

Subgrid scale models for unresolved turbulent fluxes are investigated with a focus on combustion for space propulsion applications. An extension to the gradient model is proposed, introducing a dependency on the local burning regime. The dynamic behavior of the model's coefficients is investigated, and scaling laws are studied. The discussed models are validated using a DNS database of a high-pressure turbulent fuel-rich methane-oxygen diffusion flame. The operating point and turbulence characteristics are selected to resemble those of modern combustors for space propulsion applications to support the future usage of the devised model in this context.

Large language model (LLM) is a representation of a major advancement in AI, and has been used in multiple natural language processing tasks. Nevertheless, in different business scenarios, LLM requires fine-tuning by engineers to achieve satisfactory performance, and the cost of achieving target performance and fine-tuning may not match. Based on the Baidu STI dataset, we study the upper bound of the performance that classical information retrieval methods can achieve under a specific business, and compare it with the cost and performance of the participating team based on LLM. This paper gives an insight into the potential of classical computational linguistics algorithms, and which can help decision-makers make reasonable choices for LLM and low-cost methods in business R&D.

Sign Language Recognition (SLR) aims to bridge speech-impaired and general communities by recognizing signs from given videos. Researchers still face challenges developing efficient SLR systems because of the video’s complex background, light illumination, and subject structures. Recently many researchers developed a skeleton-based sign language recognition system to overcome the subject and background variation of hand gesture signs. However, skeleton-based SLR is still under exploration due to the lack of information and annotations on hand key points. More recently, researchers included body and face information with the hand gesture for the SLR, but their performance and efficiency are unsatisfactory. We proposed a Multi-Stream Graph-based Deep Neural Network (SL-GDN) for a skeleton-based SLR system to overcome the problems. The main purpose of the proposed SL-GDN approach is to improve the efficiency and performance of the SLR system with a low computational cost based on the human body pose in the form of 2D landmark locations. In the procedure, firstly, we constructed a skeleton graph based on the selected 27 whole-body key points among 67 key points to solve the inefficiency problems. Then we proposed multi-stream SL-GDN to extract features from the whole-body skeleton graph for four streams. Finally, we concatenated the four different features and applied a classification module to refine the feature and recognize corresponding sign classes. Our data-driven and graph construction method increases the system’s flexibility and brings high generability to adapt various data samples. We used three large-scale benchmark SLR datasets to evaluate the proposed model: WLASL, AUTSL and CSL. The demonstrated performance accuracy table proved the superiority of the proposed model, and we believe this will be considered a great invention in the SLR domain.

The primary cilium, the cell’s sensory and signaling antenna, is a dynamic cellular organelle during the cell cycle. It is resorbed before cells enter mitosis and reformed after cells exit mitosis, and as such acts as a structural barrier to mitosis. Cancer cells that undergo rapid cell cycle and replication have a low ciliation rate. Therefore, promoting cilia formation and elongation and thereby reducing their rates of retraction hold the key to blocking cell entry into mitosis and slowing down cancer cell replication. To perform a comprehensive and efficient literature search on drugs that can promote ciliogenesis, we developed an intelligent process that integrates the GPT-3.5 application programming interface (API) into a PubMed scraper that we coded, enabling the large language model (LLM) to directly query articles for predefined user questions. Among the top candidates identified are two FDA-approved drugs, Alvocidib and Alisertib, that showed strong potential to induce cilia formation and elongation. We first confirmed that these drugs can effectively increase the ciliation rate and the cilia length of DAOY medulloblastoma cells. We then applied Alvocidib and Alisertib individually or in combination to DAOY medulloblastoma and A549 lung cancer cells and observed a statistically significant decrease in the number of viable cells. These results demonstrated the potential of using cilia-promoting drugs to suppress cancer cell replication. Additionally, it shows the massive benefits of integrating accessible large language models to conduct sweeping, rapid, and accurate literature searches.

A digital twin is a virtual representation that serves as the real-time digital counterpart of a physical object or system. Recent advancements have extended the concept of digital twins to humans, incorporating complex biological data such as DNA (Predictiv Care, Inc.) and immune system profiles. These sophisticated models go beyond mere pictorial representations, offering a more holistic digital reflection. However, a significant gap remains. The current human digital twin models are not capable of learning one’s preferences. In this pilot study, we introduce Diginality, a chatbot powered by HyperCLOVA X, a Large Language Model (LLM) developed by NAVER, Inc. Diginality, learns one’s preferences with custom training from data collected by interview-style questions on the user’s topic of interest. Our findings demonstrate that Diginality successfully answers one’s preferences, thereby adding a new dimension to the concept of human digital twins. This work represents a pioneering step towards creating a more comprehensive and psychologically nuanced human digital twin.

This study introduces a specialized Automatic Speech Recognition (ASR) system, leveraging the Whisper Large-v2 model, specifically adapted for radiological applications in French language. Our methodology focused on adapting the model to accurately transcribe medical terminology and diverse accents within the French language context, achieving a notable Word Error Rate (WER) of 17.121%. The research involved extensive data collection and preprocessing, utilizing a wide range of French medical audio content. The results demonstrate the system's effectiveness in transcribing complex radiological data, underscoring its potential to enhance medical documentation efficiency in French-speaking clinical settings. The discussion extends to the broader implications of this technology in healthcare, including its potential integration with Electronic Health Records (EHRs) and its utility in medical education. The study also explores future research directions, such as tailoring ASR systems to specific medical specialties and languages. Overall, this research contributes significantly to the field of medical ASR systems, presenting a robust tool for radiological transcription in the French language and paving the way for advanced technology-enhanced healthcare solutions.

Digital recruitment systems have revolutionized the hiring paradigm, imparting exceptional efficiencies and extending the reach for both employers and job seekers. This investigation scrutinizes the efficacy of classical machine learning methodologies alongside advanced large language models (LLMs) in aligning resumes with job categories. Traditional matching techniques, such as Logistic Regression, Decision Trees, Naïve Bayes, and Support Vector Machines, are constrained by the necessity of manual feature extraction, limited feature representation, and performance degradation, particularly as dataset size escalates, rendering them less suitable for large-scale applications. Conversely, LLMs like GPT-4, GPT-3, and LLAMA, adeptly process unstructured textual content, capturing nuanced language and context with greater precision. We evaluate these methodologies utilizing two datasets comprising resumes and job descriptions to ascertain accuracy, efficiency, and scalability. Our results reveal that while conventional models excel with structured data, LLMs significantly enhance the interpretation and matching of intricate textual information. This study highlights the transformative potential of LLMs in recruitment, offering insights into their application and future research avenues.

Into the frame of the French TANDEM project (Tsunamis in the Atlantic and the English ChaNnel: Definition of the Effects through numerical Modelling) Principia has been working on the development and qualification of two in-house CFD software’s: the 2D EOLE-SV (Saint Venant) model for simulation of large scale tsunami propagation from the source up to coastal scale and the 3D EOLE-NS (Navier-Stokes) model dedicated to tsunami coastal impact modelling. This paper presents a large range of test cases carried out into the frame of the project and dedicated to the validation of numerical codes in various tsunami wave conditions. The main aspects of phenomena such as wave generation, propagation and coastal impact are investigated on academic situations. A real case simulation is concerned as well, the devastating 2011 Tohoku event which is compared with in-situ data.

Fluctuations of protein three-dimensional structures and large-scale conformational transitions are crucial for the biological function of proteins and their complexes. Experimental studies of such phenomena remain very challenging and therefore molecular modeling can be a good alternative or a valuable supporting tool for the investigation of large molecular systems and long-time events. In this mini-review, we present two alternative approaches to the coarse-grained (CG) modeling of dynamic properties of protein systems. We discuss two CG representations of polypeptide chains used for Monte Carlo dynamics simulations of protein local dynamics and conformational transitions and, on other hand, highly simplified structure-based Elastic Network Models of protein flexibility. In contrast to classical Molecular Dynamics the modeling strategies discussed here allow quite accurate modeling of much larger systems and longer time dynamic phenomena. We briefly describe the main features of these models and outline some of their applications, including modeling of near-native structure fluctuations, sampling of large regions of the protein conformational space, or possible support for the structure prediction of large proteins and their complexes.

Large lithium-ion batteries (LIBs) in electric vehicles and energy storage systems demonstrate different performance and lifetime compared to small LIB cells, owing to the size effects generated by the electrical configuration and property imbalance. However, the calculation time for performing life predictions with three-dimensional (3D) cell models is undesirably long. In this paper, a lumped cell model with equivalent resistances (LER cell model) is proposed as a reduced order model of the 3D cell model, which enables accurate and fast life predictions of large LIBs. The developed LER cell model is validated via the comparisons with results of the 3D cell models by simulating a 20-Ah commercial pouch cell (NCM/graphite) and the experimental values. In addition, the LER cell models are applied to different cell types and sizes, such as a 20-Ah cylindrical cell and a 60-Ah pouch cell.

The article obtains large deviation asymptotic for sub-critical communication networks modelled as signal-interference-noise-ratio(SINR) random networks. To achieve this, we define the empirical power measure and the empirical connectivity measure, as well as prove joint large deviation principles(LDPs) for the two empirical measures on two different scales. Using the joint LDPs, we prove an Asymptotic equipartition property(AEP) for wireless telecommunication Networks modelled as the subcritical SINR random networks. Further, we prove a Local Large deviation principle(LLDP) for the sub-critical SINR random network. From the LLDPs, we prove the large deviation principle, and a classical McMillan Theorem for the stochastic SINR model processes. Note that, the LDPs for the empirical measures of this stochastic SINR random network model were derived on spaces of measures equipped with the $\tau-$ topology, and the LLDPs were deduced in the space of SINR model process without any topological limitations. We motivate the study by describing a possible anomaly detection test for SINR random networks.

Despite significant advancements in Artificial Intelligence (AI) and Large Language Models (LLMs), detecting and mitigating bias remains a critical challenge, particularly within social media platforms like X (formerly Twitter) and in addressing cyberbullying present on them. This research investigates the effectiveness of leading LLMs in generating synthetic biased and cyberbullying data and evaluates the proficiency of Transformer AI models in detecting bias and cyberbullying within both authentic and synthetic contexts. The study involves semantic analysis and feature engineering on a dataset of over 48,000 sentences related to cyberbullying collected from Twitter (before it became X). Leveraging state-of-the-art LLMs such as ChatGPT-4o, Pi AI, Claude 3 Opus, and Gemini-1.5, synthetic biased, cyberbullying, and neutral data were generated to deepen the understanding of bias in human-generated data. AI models including DeBERTa, Longformer, BigBird, HateBERT, MobileBERT, DistilBERT, BERT, RoBERTa, ELECTRA, and XLNet were initially trained to classify Twitter cyberbullying data and subsequently fine-tuned, optimized, and quantized for multilabel classification (detecting both biases and cyberbullying). The study's outcomes include a prototype of a hybrid application that combines a Bias Data Detector and a Bias Data Generator, validated through extensive testing.

A significant amount of data is available for bench-scale fire assessments. However, there is little surrounding large-scale tests and even less so for bench-to-large scale comparisons. Large-scale tests require more material, time and preparation, making them more expensive than bench-scale assessments. However, large-scale testing is an essential component to bench-scale fire testing. The bench-scale tests must be representative of both large-scale and real fire behaviour to provide an insight into how a material or product will behave in the event of a fire during its end use application. The few existing studies reviewed in this study show more guidance and data is needed, especially on a large-scale. Unfortunately, the data that is presented is done so in an inconsistent manner using various means of presentation, statistical analysis, and modelling that doesn’t show clear comparisons between bench and large scale. Currently, no bench-scale method shows good agreement with large-scale fire behaviour. Overall, there is a need for more large-scale testing and data for direct comparisons to be made.

Turbulent flows over and within forest canopies have recently been modeled with success using Large Eddy Simulations (LES). Validation exercises against experimental data suggest that models can be applied with a high degree of confidence for many applications, mechanical and physiological plant/atmosphere interaction analysis, seed or pollen dispersal, wildfire spread and firebrand transport, or investigation of causes of eddy-covariance technique bias. Long distances required for shear-induced turbulence to equilibrate, result in the widespread use of cyclic boundary conditions in LES atmospheric boundary layer studies. Vegetation drag dissipates air momentum in the atmosphere, but equilibrium is often achieved through compensatory momentum source, supplied by macro-scale pressure gradient forcing. Unfortunately, both classical Ekman balance or simple spatially-constant pressure gradient techniques for implementing this forcing have major drawbacks in the context of cyclic boundary conditions for the applications listed above. Among them, it is difficult to specify aspects of the mean velocity profile such as a specific desired wind velocity and direction at a reference height. In the present paper, we propose a new technique for capturing the effects of a large-scale pressure gradient force (LSPGF) that can be used at stand scale and enables simulation of realistic and specifiable wind fields. Several variants of this LSPGF are developed and analyzed here and validated against experimental data. Although this LSPGF technique is developed in the context of HIGRAD/FIRETEC wildfire simulations, LSPGF can be used for any LES wind modeling application aimed at generating detailed stand-scale wind fields with resolved turbulence and shear profiles consistent with vegetation structure in the boundary layer.

This paper discusses some of the most common misconceptions about large language models (LLMs), including the belief that they are sentient or conscious, that they are always accurate, and that they can replace human creativity. The paper also proposes a strategy for overcoming these misbeliefs, which involves educating the public about the capabilities and limitations of LLMs, developing guidelines for the responsible use of LLMs, and conducting more research to understand the potential impact of LLMs on society.

Disaster issue consultation is crucial for effective emergency management, requiring timely and accurate information to support decision-making processes. This paper explores the integration of large language models (LLMs) with advanced prompt engineering to enhance the performance of AI systems in disaster management. We propose a novel multi-round prompt engineering approach that guides LLMs to iteratively refine their responses based on evolving disaster scenarios. Our method leverages diverse disaster-related datasets and employs a combination of accuracy metrics and a customized GPT-4 scoring system for evaluation. We conduct comprehensive experiments across multiple LLM platforms, including Qwen2, ChatGPT, Claude, and GPT-4, comparing our approach with baseline models and the Tree of Thoughts (ToT) method. The results demonstrate significant improvements in both accuracy and response quality, highlighting the effectiveness of our method in providing contextually relevant and actionable insights. Additional analysis confirms the robustness and generalizability of our approach across various disaster scenarios, establishing it as a valuable tool for enhancing disaster management practices.

California has experienced a surge in wildfires, prompting research into contributing factors, including weather and climate conditions. This study investigates the complex, multiscale interactions between large-scale climate patterns, such as the Boreal Summer Intraseasonal Oscillation (BSISO), El Niño Southern Oscillation (ENSO), and the Pacific Decadal Oscillation (PDO), and their influence on moisture and temperature fluctuations, and wildfire dynamics in California. The combined impacts of PDO and BSISO on intraseasonal fire weather changes, the interplay between Fire Weather Index (FWI), relative humidity, vapor pressure deficit (VPD), and temperature in assessing wildfire risks, and geographical variations in the relationship between FWI and climatic factors within California are examined. The study employs a multi-pronged approach, analyzing wildfire frequency and burned areas alongside climate patterns and atmospheric conditions. Findings reveal significant variability in wildfire activity across different climate conditions, with heightened risks during specific BSISO phases, La-Niña, and cool PDO. The influence of BSISO varies depending on its interaction with PDO. Temperature, relative humidity, and VPD show strong predictive significance for wildfire risks, with significant relationships between FWI and temperature in elevated regions (correlation, r > 0.7, p ≤ 0.05) and FWI and relative humidity along the Sierra Nevada Mountains (r ≤ -0.7, p ≤ 0.05).

This article discusses the relationship between the discourse of evangelical churches and the production of conservative neoliberal political rationalities. We start from the premise that the Large Evangelical Corporations (in Brazil), operate within the neoliberal logic, and offer biographical solutions to social exclusion systems, indicating the individual and spiritual causes for poverty and suffering, while identifying the "deserving" of blessings, suggesting elevation through personal responsibility. In this turn, they merge self-entrepreneur individualism with rhetoric that interweaves elements of conservative communitarianism, such as tradition, naturalizing gender roles, and cultivating traditional family models, seen as key to rehabilitating the wider society. We hypothesize that, in the Brazilian case of the state of Rio de Janeiro, evangelical churches are central parts of this gear, acting as one of the main "resonance machines" (Connolly) of neoliberal policies, providing individualized solutions for structural systems of social exclusion, in a pedagogical, efficient and extensive manner. In this sense, this work benefits from fieldwork and interviews conducted in the state of Rio de Janeiro with its large evangelical churches.

Background: Diffuse large B-cell lymphoma is the most common type of non-Hodgkin's lymphoma. Due to the variety of clinical manifestations, accurate diagnosis can be challenging. Here, we present a case of DLBCL that was initially misdiagnosed as a hematoma, highlighting the importance of considering malignancy in the setting of non-responsive soft tissue swelling. Methods: A 76-year-old man presented to the emergency department with right periorbital swelling and ecchymosis. The symptoms first appeared after an injury and did not resolve spontaneously. The patient was taking warfarin for atrial fibrillation. A facial bone CT revealed a large irregular homogeneous mass. An extraconal hematoma was suspected based on clinical history and radiologic findings. Incision and drainage were performed, resulting in evacuation of old blood. Despite continued drainage, the patient's symptoms worsened. A comparison with the previous CT revealed an increase in the size of the lesion. As a result, a surgical excisional biopsy was performed. Results: Biopsy of the mass showed a diffuse infiltrate of lymphocytes around the tissue. The diagnosis of diffuse large B-cell lymphoma was confirmed pathologically. The patient was referred to hematology for further management. Conclusions: DLBCL is a rare disease associated with a poor prognosis. Due to the previous injury and the use of anticoagulants, the patient was initially misdiagnosed as having a hematoma. This delayed the prompt diagnosis and affected the timing of treatment initiation. We must be aware that a malignant mass is a potential cause of unresponsive soft tissue swelling.

This study examines the impact of prompt engineering on large language models (LLMs), focusing on a comparison between multitasking and single-task prompts. Specifically, we explore whether a single prompt handling multiple tasks — such as Named Entity Recognition (NER), sentiment analysis, and JSON output formatting — can achieve similar efficiency and accuracy to dedicated single-task prompts. The evaluation uses a combination of performance metrics to provide a comprehensive analysis of output quality. Experiments were conducted using a selection of open-source LLMs, including LLama3.1 8B, Qwen2 7B, Mistral 7B, Phi3 Medium, and Gemma2 9B. Results show that single-task prompts do not consistently outperform multitasking prompts, highlighting the significant influence of the model’s data and architecture on performance.

As Large Language Models (LLMs) increasingly accommodate larger inputs, context windows spanning hundreds of thousands or even millions of tokens are touted as promising for a wide array of applications. However, the potential decay in reasoning ability for larger inputs may compromise their utility. This study introduces a new benchmark called Find the Origin, which progressively tests the efficacy of LLMs on a simple intellectual task as the size of the context window increases. The test, conducted on 14 different LLMs for comparative analysis, demonstrated that reasoning ability is dependent on input size. Additionally, three independent tests were performed with the GPT-4 Turbo model to demonstrate its reasoning degradation in different contexts as input size increases.

Introduction. Intra-abdominal cystic formations represent heterogeneous pathologies with varied localization and clinical manifestation. The first challenge of a giant intra-abdominal cystic lesion is identifying the organ of origin. The clinical presentation of intra-abdominal cystic lesions varies from acute manifestations to non-specific symptoms or accidental discovery. Case presentation: A 2-year-old girl presents to the emergency unit with a fever of 38.5 Celsius, loss of appetite, and apathy. The investigations showed a gigantic intra-abdominal mass whose organ belonging could not be specified. Postoperatively, a giant mesenteric lymphangioma was evident, which was completely excised. Discussions Giant cystic formations modify the anatomical reports and become space-replacing formations, and the starting point is even more challenging to assess preoperatively. Nevertheless, the careful evaluation of the characteristics of the formation, the effect on the adjacent organs, the age of the patient, and the clinical picture can provide elements of differential diagnosis.

This paper presents results from numerical simulations validated by experimental results related to the effects of DSSI on the dynamic response of bridges. An in-service overpass was shaken using the T-Rex, a large-amplitude mobile shaker from the National Hazards Engineering Research Infrastructure (NHERI) facilities. Studies implementing Finite Element Modeling to develop time histories, response spectra, and eigenmodes were conducted in a forward-modeling problem setup. Two models were created to assess the DSSI effects on the dynamic response of the bridge. One model included elements that incorporate DSSI effects, while the other had fixed-base boundary conditions. The response from the DSSI FEM model matched the field results better than the fixed-base model, in terms of the peak response amplitudes and identified natural frequencies and modes. The influence of a series of factors, such as the soil shear wave velocity, bridge height, bridge foundation embedment depth and the corresponding rigidity, slenderness, and embedment ratios, on the bridge response is presented.

Wearable and mobile technology has advanced in leaps and bounds in the last decade with technological advances creating a role from enhancing healthy living to monitoring and treating disease. However, the discussion about the ethical use of such commercial technology in the community, especially in minors, is lacking behind. In this paper, we first summarize the major ethical concerns that arise from the usage of commercially available wearable technology in children, with a focus on smart watches, highlighting issues around the consent process, mitigation of risk and potential confidentiality and privacy issues, as well as the potential for therapeutic misconceptions when used without medical advice. Then through a relevant thought experiment we move on to outline some further ethical concerns that are connected to the use of wearables by minors, to wit the issue of informed consent in the case of minors, forcing them to live in the spotlight, and compromising their right to an open future. We conclude with the view that mitigating potential pitfalls and enhancing the benefits of wearable technology especially for minors requires brave and comprehensive moral debates.

The tradeoff between the design phase and the experimental setup is crucial to satisfy the accuracy requirement of Large Deployable Reflectors. Manufacturing errors and tolerances change the RMS of the reflecting surface and require careful calibration of the tie rod system to be able to fit into the initial design specifications. To give a possible solution to this problem, here two calibration methods, respectively for rigid and flexible ring truss support, are described. Starting from the acquired experimental data on the net nodal coordinates, the initial problem of satisfying the static equilibrium at the measured configuration is described. Then, two constrained optimization problems, for rigid and flexible ring truss support, are defined to meet RMS accuracy of the reflecting surface modifying the tie lengths. Finally, a case study to demonstrate the validity of the proposed methods is presented.

Large genomic rearrangements of PALB2 gene, particularly deletions and duplications, have been linked to hereditary breast-ovarian cancer. Our research specifically focuses on delineating the intronic breakpoints associated with rearrangements of PALB2 exon 11, which is crucial for understanding the mechanisms underlying these genomic changes in patients with hereditary breast and ovarian syndrome. By next-generation sequencing, we identified one duplication and three deletions of PALB2 exon 11, confirmed by Multiplex Ligation-Dependent Probe Amplification analysis. To assess the impact on transcription and potential splicing issues, Reverse-Transcription PCR was performed on patients' RNA. For detailed characterization of intronic breakpoints, primer walking approach and long-range PCR were implemented, followed by Sanger sequencing. Our analysis revealed a tandem duplication of 5134 base pairs (bp) mediated by AluY repeats located in introns 10 and 11, respectively. Moreover, identical deletions were identified in three unrelated patients, encompassing an approximate 8050 bp region mediated by AluSx elements. Both genomic alterations resulted in a truncated PALB2 protein due to introduction of a premature stop codon. This study underscores the remarkable instability of intronic regions flanking exon 11 of PALB2 and identifies a previously unreported hotspot involving Alu repeats with very high sequence homology in introns 10 and 11 of the gene. Our findings suggest avenues for further research, such as investigating the prevalence of similar genomic rearrangements in larger cohorts and exploring functional studies to understand how these alterations contribute to hereditary breast cancer pathogenesis.

The integration of generative AI (GAI) into the financial sector has brought about significant advancements, offering new solutions for various financial tasks. This review paper provides a comprehensive examination of recent trends and developments at the intersection of GAI and finance. By utilizing an advanced topic modeling method, BERTopic, we systematically categorize and analyze existing research to uncover predominant themes and emerging areas of interest. Our findings reveal the transformative impact of finance-specific large language models (LLMs), the innovative use of generative adversarial networks (GANs) in synthetic financial data generation, and the pressing necessity of a new regulatory framework to govern the use of GAI in the finance sector. This paper aims to provide researchers and practitioners with a structured overview of the current landscape of GAI in finance, offering insights into both the opportunities and challenges presented by these advanced technologies.

The shear stress generated by wall turbulence is the main cause of wall friction resistance in turbulent flow through pipes. This paper investigates the impact of inserting rest periods (regions of constant Reynolds number) within the pulsating operating cycle of velocity on the fully turbulent flow at large time-averaged Reynolds numbers, using the Large Eddy Simulation (LES) method. The study aims to explore the effect of increasing rest periods within pulsations on drag resistance. The dimensionless shear stress and drag reduction rates during different time periods of rest were analyzed. Numerical simulation results indicate that the pulsating velocity operation mode does not necessarily lead to drag reduction; it may even result in increased resistance. Inserting rest periods within the pulsation cycle can achieve drag reduction effects, with the maximum drag reduction rate reaching 21.8\%. Comparisons with experimental data and Direct Numerical Simulation (DNS) from the literature validate the feasibility of using the LES method for pipe pulsating operation modes.

The use of Retrieval-augmented generation (RAG) using large language models (LLMs) has shown potential for addressing issues such as hallucinations and inadequately contextualized responses. A pivotal stage in the RAG process involves a retriever for retrieving chunks based on semantic similarity with the query. In this study, we advocate for and provide experimental evidence supporting integrating and maintaining questions and answers (QA) formatted databases to improve retrieved-context representations and response quality. Our experiments evaluate our approach on benchmark RAG datasets using standard evaluation metrics and provide comparative analyses against state-of-the-art retrieval methods, showing the potential of our approach.

Artificial intelligence technologies based on deep convolutional neural networks and large lan-guage models have made significant breakthroughs in many tasks such as image recognition, target detection, semantic segmentation and natural language processing, but also face the problem of contradiction between the high computational capacity of the algorithms and the limited deployment resources. Quantization, which converts floating-point neural networks into low-bit-width integer networks, is an important and essential technique for efficient deploy-ment and cost reduction in edge computing. This paper analyze various existing quantization methods, showcase the deployment accuracy of advanced techniques, and discuss the future challenges and trends in this domain.

Takayasu arteritis is an inflammatory vasculitis that mainly impacts the large vessels including the aorta and its branches. Ocular manifestations may include retinal ischemia, optic neuropathy, cataracts, and retinal artery occlusion. We present two cases where central retinal artery occlusion was found in patients with Takayasu arteritis and wish to highlight to consider Takayasu arteritis in young patients with unexplained retinal artery occlusion.

Background: Large-cell neuroendocrine carcinoma presents significant treatment challenges owing to its rarity and limited therapeutic options. The LANCE study was designed to explore the survival benefits of incorporating atezolizumab in chemotherapy for metastatic LCNEC. Methods: In this non-randomized study, patients with metastatic LCNEC were treated with ei-ther standard chemotherapy plus atezolizumab or standard chemotherapy alone. The primary outcomes measured were progression-free survival (PFS) and overall survival (OS), with an ex-ploratory analysis of miR375 in plasma as a potential biomarker. Results: Of the 22 screened patients, 17 met the inclusion criteria and were stratified into the atezolizumab and chemotherapy (n=10) or chemotherapy alone(n=7) groups. After a median follow-up of 23.3 months, the 12-month survival rate for the atezolizumab group was 57.1% (95% CI: 32.6%-100%) and 45.7% at 24 months (95% CI: 22.4%-93.2%). In contrast, the chemotherapy group had a 14.3% survival rate at both 12 and 24 months (95% CI: 2.33%-87.7%). The log-rank test for survival showed a statistically significant difference in the atezolizumab-treated group (chi-squared = 4, df = 1, p = 0.04). A non-significant correlation was found between disease pro-gression and elevated levels of miR375. Conclusions: This pilot study suggests that atezolizumab combined with chemotherapy may be effective in treating metastatic LCNEC. The non-significant correlation between plasma miR375 levels and disease progression suggests its biomarker potential, warranting further validation research.

Punica granatum (PG) or Pomegranate is a bush, having many trunks and often get taller 1.8 to 4.7m long and their leaves are shiny and fleshy, approximately 7 to 8cm in length. Traditionally, it is used in diarrhea and dysentery due to its strong constricting property. Punica Granatum covering used in dysentery, diarrhea, piles, bronchitis. It is also used to minimize the risk of heart disease. The peel extract of Punica Granatum contain methanol and their dried peels has antidiarrheal venture, also help in wound healing activity. Peels of the pomegranate also holds around 60% of the fruit and they contain different types of ingredients including ellagitannins and proanthocyanidin compounds and minerals such as calcium, phosphorus, magnesium, potassium, flavonoids and sodium. Some animals were found in different local areas with both acute and chronic diarrhea. The condition of these animals were quite miserable because of dehydration and loss of electrolytes in the body cells of the animals. They were unable to move due to seriousness of the disease. 8 grams of PG Peel Powder mixed with 1 liter of water, 2 g of PG Peel Powder mixed with 250 ml water and 8 g of PG Peel Powder mixed with 1 liter of water were administrated orally by using sterilized bottles to the selected three cases. 3 days after the administration of different doses, all cases of diarrheal animals show a significant recovery. The condition of the animals seem to be healthier. Before administration of the pomegranate peel powder extract, all three cases had weak body structure due to diarrhea. Watery stool that may be yellow in color and mucus was also evident in the stool, due to this reason the patient calf was weak and depressed. In sometimes they also show unconsciousness and blood glucose levels seem to be reduced from the normal level.

This paper deals with the analysis of the cognitive functions of the cerebral cortex or cortex in the evolutionary context. The triune brain model is presented given its notorious popularity in fields that are not strictly neuroscientific, and its limitations and inaccuracies are also shown. Although it is an outdated theory, this one represented the passage to a new era, because it really was the model of the movement towards understanding the brain in evolutionary terms. Next, the paper illustrates the different areas of the cortex and their functions, as well as the evolution of the brain and the cortex in vertebrates. To date, the expansion of the cortex with the amniotes was placed 300 million years ago, although recent studies place it in the lampreys more than 500 million years ago. Evolutionary changes in mammalian brain and cortex size are discussed, showing the evolution from three to six cell layers and trends in different groups of mammals. Finally, the most recent contributions in the field of research are illustrated, such as the role of neurogenesis and the evolution of the connectivity of the cortex.

Swirling coal flames are encountered in swirl burners of boiler furnaces and cyclone coal combustors, and were studied using large-eddy simulation (LES) in reported references. However, the structures of swirling coal flames remain to be further studied. In the present paper LES of swirling coal flames is conducted for two combustors using two gas combustion models. The statistical simulation results are accessed by experiments and the instantaneous simulation results are used to analyze the specific features of flame structures of four cases. It is seen that different inlet flow conditions and different secondary-air flow rate lead to different flame structures

Cycas (family Cycadaceae), spread throughout tropical and subtropical regions, is crucial in con-servation biology. Due to the subtle morphological variations between species, a solid spe-cies-level phylogeny for Cycas is lacking. Because of the rapid progress in high-throughput se-quencing technology, it has become feasible to acquire complete chloroplast (cp) genome se-quences, which provide a molecular foundation for phylogenetic research. In the present study, we employed next-generation sequencing technology to assemble and analyze the chloroplast genomes of six Cycas plants, including their genome structure, GC content, and nucleotide di-versity. The Cycas chloroplast genome spans 162,038 to 162,159 bp and contains 131 genes, in-cluding 86 protein-coding genes, 37 transfer RNA (tRNA) genes, and 8 ribosomal RNA (rRNA) genes. Through a comparative analysis, we found that the chloroplast genome of Cycas was highly conserved, as indicated by the contraction and expansion of the inverted repeat (IR) re-gions and sequence polymorphisms. In addition, several non-coding sites (psbK-psbI, petN-psbM, trnE-UUC-psbD, ndhC-trnM-CAU, and rpl32-trnP-GGG) showed significant varia-tion. The utilization of phylogenetic analysis relying on protein-coding genes has substantiated the division of Cycas primarily into four groups. The application of these findings will prove valuable in evaluating genetic diversity and the phylogenetic connections among closely related species. Moreover, it will provide essential support for the advancement of wild germplasm re-sources.

Although ChatGPT promises wide-ranging applications, there is a concern that it is politically biased; in particular, it has a left-libertarian orientation. Nevertheless, in light of recent trends in attempts to reduce such biases, this study re-evaluated the political biases of ChatGPT using political orientation tests and the application programming interface. Moreover, the effects of the languages used in the system as well as gender and race settings were evaluated. The results indicated that ChatGPT had less political bias than previously thought; however, they did not entirely discount the political bias. The languages used in the system and the gender and race settings may induce political biases. These findings enhance our understanding of the political biases of ChatGPT and may be useful for bias evaluation and designing ChatGPT’s operational strategy.

The paper shows an analysis of the large-scale distribution of galaxy spin directions of 739,286 galaxies imaged by DES. The distribution of the spin directions of the galaxies exhibits a large-scale dipole axis. Comparison of the location of the dipole axis to a similar analysis with data from SDSS, Pan-STARRS, and DESI Legacy Survey shows that all sky surveys exhibit dipole axes well within 1$\sigma$ error from each other. While non-random distribution is unexpected, the findings are consistent across all sky surveys, regardless of the telescope or whether the data were annotated manually or automatically. Possible errors that can lead to the observation are discussed. The paper also discusses previous studies showing opposite conclusions, and analyzes the decisions that led to these results. Although the observation is provocative, and further research will be required, the existing evidence justifies to consider the contention that galaxy spin directions as observed from Earth are not necessarily randomly distributed. Possible explanations can be related to mature cosmological theories, but also to the internal structure of galaxies.

The forming of large rivers are the integral consequences of the deep earth process and the surface. In contrast to the hot topics for rivers related to orogenic domains, rift-related large rivers are largely ignored especially in deep time studies. The Cenozoic East Asia margin provides very good opportunity to observe this kind of rivers. It has been believed that basin-and-swell physiography dominated the East Asia margin and impeded the forming of large rives in the early Cenozoic. In this paper, we combined provenance analysis of East China Sea Basin, where is a crucial place to trace the river evolution in East Asia margin, and regional geologic constraints to reveal drainage reorganizations. Detrital zircon U-Pb ages from the Early Eocene sediments of the East China Sea Basin are firstly reported. Our results together with literature data demonstrate that regional provenance changes occurred at the middle Eocene from one singe age peak at ~110 Ma of proximal sources to multiple age spectrum derived from far inland. Source to sink analysis indicated that the North China Block and Korea Peninsular provided the most detritus. Sedimentation and tectonic features of rift basins in the potential source areas indicated that rivers flowed into Bohai Basin and Jianghan Basin cannot provide terrigenous clasts for the lower reaches in the Eocene. Contrastingly, the dominantly fluvial sediments across the Subei-South Yellow Sea Basin suggested external river system and a bypassing region since the middle Eocene, coinciding with provenance change in ECSB. All these demonstrated that a large river (East Asia River) established in east Asia margin in the middle Eocene and flowed southwestward approximately 1500km to the sea in southern ECSB. This river might last to the middle Miocene. The deep earth processes driven by Izanagi-Pacific ridge subduction resulted in the overfilled stage of Subei-South Yellow Sea Basin and the post-rift subsidence in west depression of ECSB, and thus facilitated the initiation of the EAR. Our finding shed new light on the evolving landscape in East Asia and showed how subduction of deep earth process controlled the initiation of rift-related large rivers.

A robust hydrological assessment is challenging in regions where human interference, within all aspects of the hydrological system, significantly alters the flow regime of rivers. The challenge was to extend a large-scale water resources model, GWAVA, to better represent water resources without increasing the model complexity. A groundwater and a regulated reservoir routine were incorporated into GWAVA using modifications of the existing AMBHAS-1D and Hanasaki methodologies, respectively. The groundwater routine can be varied in complexity when sufficient input data is available but fundamentally is driven by three input parameters. The reservoir routine was extended to account for the presence of large, regulated reservoirs using two calibratable parameters. The additional groundwater processes and reservoir regulation was tested in two highly anthropogenically influenced basins in India: the Cauvery and Narmada. The inclusion of the revised groundwater routine improved the simulation of streamflow in the headwater catchments and was successful in improving the representation of the baseflow component. In addition, the model was able to produce a time series of daily groundwater levels, recharge to groundwater and groundwater abstraction. The regulated reservoir routine improved the simulation of streamflow in catchments downstream of major reservoirs, where the streamflow was largely reflective of reservoir releases, when calibrated using downstream observed streamflow records. The model was able to provide a more robust representation of the annual volume and daily outflow released from the major reservoirs and simulate the major reservoir storages adequately. The addition of one-dimensional groundwater processes and a regulated reservoir routine proved successful in improving the model performance and traceability of water balance components, without excessively increasing the model complexity and input data requirements.

The mean flow and turbulence statistics of the flow through a simplified urban environment, which is an active research area in order to improve the knowledge of turbulent flow in cities, is investigated. This is useful for civil engineering, pedestrian comfort and for health concerns caused by pollutant spreading. In this work, we provide analysis of the turbulence statistics obtained from well-resolved large-eddy simulations (LES). A detailed analysis of this database reveals the impact of the geometry of the urban array on the flow characteristics and provides for a good description of the turbulent features of the flow within a simplified urban environment. The most prominent features of this complex flow include coherent vortical structures such as the so-called arch vortex, the horseshoe vortex and the roof vortex. These structures of the flow have been identified by an analysis of the turbulence statistics. The influence of the geometry of the urban environment (and particularly the street width and the building height) on the overall flow behavior have also been studied. Finally, the well-resolved LES results were compared with the experimental database from Monnier et al. to discuss differences and similarities between the respective urban configurations.

This study explores how to delineate Russian cities in order to make them comparable on the world scale. In doing so we introduce the concept of large urban regions (LUR) applicable to the Russian urban context. This research is motivated by a principal research question: how to construct a statistical urban delineation, which would allow first, to demonstrate integration of cities into globalization, and second, to make global urban comparative research. Previous studies on urban delineation in Russia have focused almost exclusively on functional urban areas, which have substantial limitations and are not suitable for global urban comparisons. Addressing this research gap, we propose a new definition of Large Urban Regions (LUR). In doing so, first, we introduce the context of Russian cities (2), then we discuss existing Russian urban concepts (3), and justify a need for a new urban delineation (4). Afterwards, we present a general method to delineate Large Urban Regions in Russian context (5.1), and illustrate it in the two case studies of St. Petersburg (polycentric region) and Samara (monocentric region) (5.2). In the last part (6), we discuss the 10 the largest urban regions in Russia and describe a constructed database including all Russian LURs.

The attributes of electroplating as a low-cost, simple, scalable and manufacturable semiconductor deposition technique for the fabrication of large-area and nanotechnology-based device applications are discussed. These strengths of electrodeposition are buttressed experimentally using techniques such as X-ray diffraction, Ultraviolet-visible spectroscopy, Scanning electron microscopy, Atomic force microscopy, Energy-dispersive X-ray spectroscopy and photoelectrochemical cell studies. Based on the structural, morphological, compositional optical, and electronic properties evaluated results, it is evident that electroplating possesses the capabilities of producing high-quality semiconductors usable in producing excellent devices. In this paper, we will describe the progress of electroplating technique mainly for the deposition of semiconductor thin film materials, their treatment processes and fabrication of solar cells.

The performance in terms of minimal Bayes’ error probability for detection of a high-dimensional random tensor is a fundamental under-studied difficult problem. In this work, we consider two Signal to Noise Ratio (SNR)-based detection problems of interest. Under the alternative hypothesis, i.e., for a non-zero SNR, the observed signals are either a noisy rank-R tensor admitting a Q-order Canonical Polyadic Decomposition (CPD) with large factors of size Nq  R, i.e, for 1  q  Q, where R, Nq ! ¥ with R1/q/Nq converge towards a finite constant or a noisy tensor admitting TucKer Decomposition (TKD) of multilinear (M1, . . . ,MQ)-rank with large factors of size Nq  Mq, i.e, for 1  q  Q, where Nq,Mq ! ¥ with Mq/Nq converge towards a finite constant. The detection of the random entries (coefficients) of the core tensor in the CPD/TKD is hard to study since the exact derivation of the error probability is mathematically intractable. To circumvent this technical difficulty, the Chernoff Upper Bound (CUB) for larger SNR and the Fisher information at low SNR are derived and studied, based on information geometry theory. The tightest CUB is reached for the value minimizing the error exponent, denoted by s?. In general, due to the asymmetry of the s-divergence, the Bhattacharyya Upper Bound (BUB) (that is, the Chernoff Information calculated at s? = 1/2) can not solve this problem effectively. As a consequence, we rely on a costly numerical optimization strategy to find s?. However, thanks to powerful random matrix theory tools, a simple analytical expression of s? is provided with respect to the Signal to Noise Ratio (SNR) in the two schemes considered. A main conclusion of this work is that the BUB is the tightest bound at low SNRs. This property is, however, no longer true for higher SNRs.

With increasing global attention on carbon emission control, the carbon market has become an important tool for carbon emission regulation. As an important part of energy trading, the electricity market is closely related to the carbon market. Given the relative maturity of the EU carbon market and electricity market, this paper identifies key points of market volatility through the Coupling coordination model analysis, and uses the GPT large model analysis to analyze policy factors that have a greater impact on carbon market and electricity market transactions. The research results show that market-regulating policies and expectation management policies have the greatest impact on the degree of coupling between the carbon and electricity markets. The impact of secondary policies such as international cooperation policies and structural adjustment policies is relatively insignificant. Although the development of China's carbon market faces many challenges, reasonable policy interventions are expected to achieve significant results in the domestic carbon market. Finally, this paper also draws on analogy analysis to draw corresponding implications for China from the EU's carbon-electricity coupling policy promulgation.

Securing funding is a critical yet complex task for organizations and individuals. This study presents an innovative chatbot designed to streamline the process using advanced Natural Language Processing (NLP) techniques and, specifically, a Retrieval-Augmented Generation (RAG) pipeline optimized for real-world applications. Our chatbot assists users in identifying suitable public tenders for financial support through natural language queries and a comprehensive public data database. The chatbot operates in a two-stage interaction model, initially providing summarized tender information for exploratory brainstorming, followed by detailed data upon user selection. A custom filtering mechanism ensures that the user interface elements responsible for swapping the interaction stages are consistent with the responses generated by the conversational agent. Human-evaluation tests demonstrated an average accuracy of 90.4% in document retrieval, with an average of 2.11 interactions required to find a specific tender. User satisfaction, rated on a scale of 1 to 5, averaged 3.14 (±1.73), indicating generally positive user experience with room for improvement. This approach addresses challenges of relevance, accuracy, and conversational flow, resulting in a reliable chatbot that simplifies the process of finding funding opportunities.

In the rapidly evolving landscape of AI-supported programming education, there's a growing interest in leveraging such tools to support students helping about what good code makes out. This study investigates the impact of ChatGPT on code quality among part-time undergraduate students enrolled in introductory Java programming courses. With no prior Java experience, students from two separate groups completed identical programming exercises emphasizing coding conventions and code quality. Utilizing static code analysis tools, we assessed adherence to a common coding convention ruleset and calculated cyclomatic and cognitive complexity metrics for submitted code. Our comparative analysis highlights significant improvements in code quality for the ChatGPT-assisted group (treatment group), with marked reductions in rule violations and both cyclomatic and cognitive complexities. Specifically, the treatment group demonstrated greater adherence to coding standards, with fewer violations across several rules, and produced code with lower complexity. These results suggest that ChatGPT can be a valuable tool in programming education, aiding students in writing cleaner, less complex code and better adhering to coding conventions. However, the study's limitations, including the small sample size and the novice status of the participants, necessitate further research with larger, more diverse populations and in different educational contexts.

In the context of rapid urbanization, the spatiotemporal patterns of socio-economic exposed to high seismic hazard in large cities lacks quantitative assessment, thus, it is difficult to understand the evolution of earthquake catastrophe risk there. In this paper, exposure change rate, and absolute and relatively exposure indices were proposed to quantitatively study the changes of spatiotemporal exposure to high seismic hazard in urban areas of large cities (i.e., permanent residents in urban areas more than 1 million) in Chinese mainland, using a set of multi-temporal population, GDP, and built-up datasets and a seismic hazard zoning map. The results show that, 64 out of 78 large cities are entirely or partially located in high seismically hazardous zones. Although the urban areas of large cities in high seismically hazardous zones (HSHUAs) cover only 1.73% (167177 km²) of mainland China, their share of population, GDP, and built-up land amounted to 18.14%, 30.24% and 24.55%, respectively, in 2015. Unprecedented urbanization has led to a dramatic exposure growth in HSHUAs between 1990 and 2015. The population, GDP, and built-up land in HSHUAs increased by 110 million people (with an average annual growth rate of 2.37%), 273.71 trillion USD(average annual growth rate of 11.15%), and 10593 km2 (average annual growth rate of 2.02%), respectively. HSHUAs experienced much faster population and GDP growth than the whole country, while relatively slower growth in built-up land. Exposure growth in HSHUAs also was varied for different types of large cities. Megacities had a higher proportion of increased GDP and population in HSHUAs, induced by industry and population agglomeration. Whereas, Type II large cities had a higher proportion of increased built-up land in HSHUAs, driven mainly by urban sprawl. Our analysis indicates that, for large cities under rapid urbanization, it is an urgent need to strengthen resilient urban planning, renewal and construction.

Today, LLMs are not good at personalization providing recommendation. They advise physicians and financial advisors to ask professionals in respective fields for help, even having user information available. Answering questions of software professionals, LLM needs to deliver in-depth answers with codes or algorithms, whereas for professionals in other fields would need definitions and main concepts. The intent of this chapter is to make LLM answer tailored to the needs of users, taking into account available information about them. To do that, we need to generalize available information about a person like her health record, maintaining the privacy of this person. We rely on meta-learning techniques to design a LLM prompt to produce a personalization prompt to obtain a suitable relevant information. Such “meta-prompt” is produced by generalization operation applied to available documents for the user. These documents need to be de-identified so that they are sufficient for personalization on one hand and will maintain user privacy on the other hand. The second neuro-symbolic technique to support personalization is abductive reasoning, acting in parallel to LLM fine-tuning. Traditional recommendation and personalization techniques as well as modern, deep learning – based are presented, and the comparison is drawn to the proposed approach. We also share the evaluation and comparative analyses of these approaches. We consider an example for how to build personalization LLM systems coming from Langchain platform. We will explore how to construct chains to form a personalization profile for a user and apply it to user search and recommendation requests.