Articles

This review provides comparative analysis of GPT-4, BERT (bidirectional encoder representations from transformers), Gemini, and DeepSeek large language models (LLM), focusing architectures, training methodologies, and real-world applications. The primary research question is: How do these models differ in design, strengths, limitations, and potential areas for enhancement? By addressing this question, the study aims to provide insights into the trade-offs and future directions for optimizing LLM performance and deployment. The analysis reveals GPT-4 excels in natural language generation and complex reasoning, supporting up to 128K tokens with moderate latency and higher costs making it effective for conversational artificial intelligence (AI). BERT excels bidirectional contextual understanding with smaller computational overhead and broad open-source adoption, effective for text classification. Gemini demonstrates superior multimodal integration processing text, image, audio, and code with context lengths up to 1M tokens, enabling cross-domain adaptability. DeepSeek excels in specialized domains like finance and programming, is optimized for efficiency and supports extended context windows exceeding 200K tokens. However, all models face challenges related to computational cost, hallucinations, and ethical concerns, necessitating further improvements. Despite advancements, LLMs continue to grapple with issues such as data bias, model interpretability, and responsible AI deployment. Future research should focus on hybrid model approaches, domain-specific fine-tuning, and transparency to mitigate risks while maximizing the transformative potential of LLMs in real-world applications.

Significant advances have been achieved in machine translation (MT) in recent times, particularly state of the art (SOTA) models for languages like English and Indian having distinct grammatical structures and limited monolingual training data. This paper analyses various recent state-of-the-art variants of large language models (LLMs) and neural machine translation (NMT) for Indian languages in comparison to statistical machine translation (SMT). It tackles key questions, such as idiomatic expressions, morphologically complex grammar or the scarceness of parallel corpora. Furthermore, it studies bytewise BPE, compares translation models in terms of BLEU scores using separate and shared-vocabulary representation with copy actions between the BPE translations, and analyses how multitask learning (Caruana (1997)) and attention mechanisms can contribute to the quality of translation. In summary, it provides directions for future work by suggesting new avenues of research including better curated datasets, more efficient approaches for lowresource languages and culturally aware translations.

New students often struggle when short articles clash with thick textbooks. Still, even though large language models offer some teaching support, standard online setups lack focused accuracy - sometimes making things up - and risk user data control. Here comes an idea: build a tightly tested, self- contained system that aligns learning materials automatically without needing the internet, keeping information private by design. One look at two setups shows how they handle local reasoning differently. Instead of using both encoder and decoder parts, one system skips the encoder entirely. That simpler design grabs full context through ChromaDB without shrinking the data first. Meanwhile, the older type crunches input down, losing meaning along the way. Even though it runs fast - just under a second - errors pop up often, four out of five responses drifting off course. On the flip side, the new method builds correct code nearly every time, adds clear explanations tied to lesson goals, yet takes more than fourteen seconds to reply. Slower? Yes. More accurate? Clearly. What stands out is how compressed models running locally can still catch up in understanding classroom content. Another key point emerges: building tutors powered by artificial intelligence (AI) becomes safer when data never leaves the device and outside services are not needed at all.

The increasing prevalence of toxic content on social media, conveyed through text, images, and videos, poses significant challenges for automated content moderation systems. Although prior studies have reported promising results in unimodal and bimodal settings, they often fail to capture implicit and contextual toxicity emerging from interactions across multiple modalities, particularly in non-English environments. This paper proposes WVisionBERT-VL, an end-to-end multimodal framework for toxicity detection that integrates text, image, and video modalities within a unified architecture. The proposed model incorporates modality-specific encoders, bidirectional multi-head cross attention (BMHCA) for cross-modal synchronization, and an adaptive fusion gate to dynamically balance modality contributions. A balanced multimodal dataset is constructed from social media platforms, including X, Instagram, and TikTok, and refined using a model-based labeling strategy with limited human-in-the-loop validation. Experimental results on a custom Indonesian dataset demonstrate strong in-domain performance, achieving an accuracy of 94.12%, a macro-F1 of 0.9407, and a receiver operating characteristic - area under the curve (ROC-AUC) of 0.9721, with robustness further validated through five-fold cross-validation. Cross-dataset evaluation highlights challenges related to domain shift, underscoring the need for future research on robust and domain-adaptive multimodal toxicity detection.

Social media platforms have greatly accelerated the spread of news, but this rapid information flow also amplifies the risk of misinformation. Traditional automatic detection methods that rely solely on textual features often struggle with nuanced, emerging content. This paper present a novel pipeline that verifies the factuality of social news by clustering related articles into events and using large language models (LLMs) to extract and verify claims against trusted news sources. The approach groups social-media posts and mainstream reports on the same event, extracts atomic claims with a model like GPT-4, and checks each claim’s truthfulness by comparing it to the cluster’s reliable news. This pipeline was evaluated on a newly constructed Vietnamese news dataset of 1,765 articles (including 723 social-media items), manually annotating claims as true or false. The LLM-based method achieved high accuracy (≈88.9% F1-score on claim verification and 92.1% F1-score on overall news verification). These results demonstrate that carefully prompted LLMs, combined with event-level clustering of evidence, can outperform traditional methods (e.g., bidirectional encoder representations from transformers (BERT)-based classifiers) in verifying news. The paper discusses advantages of clustering over simple retrieval, scalability considerations for LLMs, and prospects for multilingual and knowledge-enhanced verification.

This study examines the reliability of an image-native multimodal AI system for automated scoring of handwritten responses to Advanced Placement (AP) Calculus free-response items without requiring optical character recognition (OCR) preprocessing. Using inter-rater agreement indices and test–retest reliability analyses, we found substantial to almost perfect agreement between artificial intelligence (AI)-generated scores and calibrated human ratings, as well as almost perfect stability across repeated scoring sessions. These results suggest that the observed reliability of the AI scoring system warrants further investigation of validity-related evidence and inferences. As a practical implication for assessment practice, we propose a human-in-the-loop teacher–AI collaborative (TAC) framework in which automated scoring operates under teacher oversight. Taken together, these findings provide initial evidence of reliability supporting the responsible use of AI-based scoring as a measurement instrument in high-stakes educational assessment.

The present study aims to investigate the impact of integrating artificial intelligence (AI) technologies into teaching Russian as a foreign language (RFL) from the perspective of educators. Employing a mixed-methods research design, the study utilized several methodologies, including a teacher survey, an analytical-descriptive approach to data interpretation, and the development and evaluation of AI-based interventions. The study sample comprised 120 RFL instructors from three public universities in Kazakhstan. Preliminary findings revealed a considerable awareness among teachers regarding various AI technologies, such as chatbots, voice assistants, the ChatGPT neural network, educational platforms, gaming applications, and task design tools. Nonetheless, the practical utilization of these technologies varied significantly, with only a subset of teachers incorporating them into their regular teaching practices. The study culminated in the development of a conceptual framework for AI-driven educational interventions, incorporating platforms such as Coursera, Moodle, Open EdX, and eFront; game-based applications including Duolingo, Talk2Russia, and Russian Verbs Pro; and task creation tools such as Kahoot! and Quizlet. Following the integration of these interventions into the curriculum, post-implementation evaluations indicated that teachers generally perceived the tools as effective, with the average effectiveness rating surpassing 4.0 out of 5.0 across all assessed categories. The findings of this study have practical applicability; they can be used to enhance professional development programs for teachers of the Russian language and to formulate strategies for the integration of AI technologies into language education within the Central Asian region.

This paper presents an adaptive retrieval-augmented generation (RAG) framework for enterprise knowledge systems that combines multi-source ingestion, semantic indexing with Hugging Face embeddings and Facebook artificial intelligence similarity search (FAISS), metadata-aware retrieval, and grounded large language model generation. The research addresses a persistent enterprise gap: critical knowledge is distributed across documentation, tickets, code repositories, and collaboration tools, while static keyword search and periodically retrained language models cannot keep pace with rapidly changing operational data. The proposed approach contributes a privacy-preserving architecture, a retrieval-and-feedback loop that improves ranking quality over time, and a unified workflow that links evidence retrieval to solution recommendation. In an evaluation over a 1.2 million-document corpus and a six-week pilot, the framework improved Precision@10 from 0.58 to 0.81, reduced documentation retrieval latency from 45.6 s to 12.3 s, and shortened average bug-resolution time from 18.4 h to 7.2 h. These findings indicate that enterprise RAG can materially improve troubleshooting speed, knowledge reuse, and decision support while maintaining stronger control over sensitive organizational data. The broader implication is that adaptive, governed RAG systems can serve as a practical foundation for future enterprise artificial intelligence (AI) assistants, analytics platforms, and compliance-aware decision workflows.

This paper presents a survey of retrieval algorithms used in advertising recommendation and organic content recommendation systems. Modern digital platforms rely on retrieval-based models to efficiently match users with relevant advertisements or personalized content. This survey reviews key techniques including inverted index methods, collaborative filtering, content-based filtering, hybrid recommendation models, and the two-tower neural network architecture widely used in large-scale recommendation systems. The paper compares the objectives, data utilization strategies, and evaluation metrics of ad targeting and organic retrieval systems. Practical challenges such as cold-start problems, data quality, scalability, and privacy considerations are also discussed. This survey further highlights the growing connection between industrial recommendation pipelines and emerging retrieval mechanisms used in large language model (LLM) systems. This survey provides insights into the design principles of modern retrieval systems and outlines future research directions at the intersection of recommendation systems and LLM.

This survey characterizes the contributions of large language models (LLMs) to technology enhanced learning by relating their capabilities to actual educational functions, making comparisons with traditional models of language. The contributions for this study are: i) introduce an education centered taxonomy that classifies LLM use by four key functions personalization and adaptivity, assessment and evaluation, profiling and prediction, and intelligent tutoring with illustrations from deployed systems and tools; ii) give a domain-based comparison of where LLMs outperform traditional models (sentiment analysis with sarcasm, context-aware question answering, and abstractive summarization) and why those advantages will mean something to e-learning practice; iii) synthesize six cross-cutting risks, including computational cost/carbon, privacy, bias and hallucination, labor displacement, interpretability, and the limits of human-like judgment, and provide practical design/research implications; and iv) report on a transparent review protocol that got the initial corpus down to 50 key articles, allowing for modifications and future updates from other interested researchers. In sum, the discussion about LLMs in education has been pushed past the broad strokes to a situation where there is a comprehensive vocabulary for what LLMs can do, and how they may or may not responsibly improve learning experiences, educator workflows, and systems/learning design in e-learning.

This study examines the elevated prevalence of type 2 diabetes mellitus (T2DM) in Saudi Arabia by integrating a large-scale, regionally specific dataset from the Saudi Ministry of Health (100,000 patient records). The primary contribution is this Saudi-focused data integration—addressing a critical gap in prior studies that rely on global datasets (e.g., PIMA, UK Biobank)—and a comparative evaluation of traditional machine learning (ML) models (random forest (RF), adaptive boosting (AdaBoost)) and few shot large language models (LLMs) for early risk detection. Key features include age, body mass index (BMI), HbA1c, blood glucose, hypertension, and smoking history. Class imbalance (9% diabetic) was resolved using synthetic minority over-sampling technique with edited nearest neighbors (SMOTEENN). Ensemble ML models achieved up to 97% accuracy, while few-shot LLM reached 98% across accuracy, precision, recall, and F1-score (area under the curve (AUC) 0.99). Feature importance confirmed that HbA1c and glucose were the top predictors. This regionally tailored, high performing framework enables early detection of T2DM and culturally sensitive interventions in high-prevalence settings such as Saudi Arabia.

Large language models (LLMs) have recently expanded the scope of automation across many application domains. In geographic information systems (GIS), however, many tasks still require specialized expertise and remain difficult for non-expert users. Recent studies have explored LLM-based geospatial analysis under a single-agent paradigm, but these early systems remain limited by weak coordination, limited error recovery, and dependence on proprietary artifacts. This study proposes multi-agent autonomous geospatial artificial intelligence (MA-GeoAI), a multi-agent architecture in which the planner, coder, validator, debugger, and knowledge agents collaborate through the LangGraph framework. The framework was evaluated on three case studies: population exposure assessment, mobility pattern analysis, and county-level mortality modeling. Unlike general-purpose multi-agent LLM frameworks, MA-GeoAI embeds spatial semantics, coordinate reference system (CRS) consistency checks, geometry validation, and operation-aware coordination directly into the control loop. Across repeated runs, all evaluated systems completed the controlled artifact contract; therefore, the analysis focuses on auditability, runtime, fallback behavior, and reproducibility rather than binary task-completion superiority.

Arabic sentiment analysis (SA) faces significant challenges due to the language’s morphological richness and dialectal diversity. This study introduces a novel hybrid large language model-graph neural network (LLM-GNN) framework designed to address these challenges. The proposed model integrates the contextual understanding of AraBERT v2 with the structural learning capability of a graph convolutional network (GCN). It constructs a graph of sentences using cosine similarity, allowing the GCN to capture crucial inter-sentence semantic dependencies often missed by sequential models. Findings: The model is evaluated on a publicly available Arabic 100k Reviews dataset consisting of authentic user-generated Arabic reviews balanced across positive, negative, and mixed sentiment classes. The results demonstrate that the proposed LLM-GNN model performed better as compared to the baseline models, including fine-tuned AraBERT, AraBERT-BiLSTM, AraBERT-MLP, and multilingual BERT. The hybrid model achieves an overall accuracy of 66.8% and a F1-score of 66.55%, with an improvement of 7.6% and 4.4%, respectively. The model demonstrated stable convergence from the first training epoch. Research limitations/implications: The graph construction is performed at the mini batch level, which restricts the modeling of global semantic relationships across the entire corpus. The results show that the hybrid model identifies subtle sentiment cues that sequential models frequently miss by fusing relational graph reasoning with contextual embeddings. By effectively identifying subtle sentiment cues, the hybrid model can significantly enhance the accuracy of real-world applications such as social media monitoring and customer review analysis for Arabic content.

Sentiment analysis is a core task in natural language processing with broad ap plications in social media monitoring, customer feedback mining, and market research. Although pre-trained language models (e.g., BERT) achieve strong performance, they typically rely on task-specific fine-tuning and substantial la beled data. Recent large language models (LLMs) enable a different paradigm via in-context learning. This paper presents a systematic empirical study investi gating chain-of-thought sentiment (CoT-Sent), a prompting framework that uses structured CoT reasoning to improve classification accuracy. We evaluate CoT Sent on four benchmark datasets in English and Chinese, comparing multiple representative LLMs (GPT-4, Claude-3, Gemini, Qwen-2.5) under zero-shot set tings. Across datasets, CoT-Sent improves average accuracy by 2.5% over zero shot baselines. Crucially, unlike prior work which provides a broad performance overview without analyzing deployment costs or multi-language generalization, we focus on the cost-latency-accuracy trade-offs, and demonstrate CoT-Sent’s superior cross-lingual transfer (English-to-Chinese) with detailed cost analysis. We provide a comprehensive three-dimensional analysis of accuracy, cost, and latency, offering actionable deployment strategies for resource-constrained environments.

Large language models (LLMs) have transformed natural language processing (NLP), demonstrating exceptional proficiency in tasks such as text generation, translation, and summarization. However, LLMs are prone to generating biased, inaccurate, or contextually irrelevant outputs, posing significant risks in high-stakes domains such as healthcare, legal reasoning, and engineering. This paper systematically investigates the role of prompt engineering as a solution to these challenges. By strategically designing inputs, prompt engineering enhances LLM performance, yielding more accurate, contextually relevant, and ethically aligned outputs. Advanced techniques, including chain-of-thought (CoT) prompting and retrieval augmented generation (RAG), are examined for their ability to improve reasoning capabilities, reduce errors, and mitigate bias. CoT prompting facilitates structured, stepwise reasoning, while RAG incorporates real-time data, ensuring output accuracy in rapidly evolving fields. In addition, we present a novel comparative perspective on these techniques, highlighting their distinct strengths and limitations across specialized applications such as healthcare diagnostics and scientific data extraction. The findings demonstrate that sophisticated prompt engineering significantly elevates the reliability and precision of LLM outputs, while addressing critical ethical concerns such as data privacy, bias, and hallucination. These insights underscore the necessity of advanced prompt design in optimizing LLMs for high-impact applications, ensuring both performance and ethical integrity.