Search Results (191)

Speaking proficiency remains one of the most challenging skills for learners of English as a Foreign Language (EFL), particularly in contexts where sustained spoken interaction is limited. This systematic review synthesises 36 empirical studies (2015–2025) identified through a PRISMA-guided Scopus search to examine how artificial intelligence (AI)-mediated instruction supports EFL speaking development. The included studies were analysed according to AI modality, pedagogical integration, instructional input characteristics, and linguistic and affective outcomes. Findings indicate that AI tools—such as chatbots, automatic speech recognition systems, and large language models—consistently support affective outcomes, including reduced speaking anxiety and increased willingness to communicate. Improvements in fluency, pronunciation, and accuracy were frequently reported, particularly when AI tools were embedded within task-based and pedagogically structured instructional designs. However, evidence for sustained development of higher-order communicative competence was more variable. The review proposes a mediated input framework conceptualising AI as a design-sensitive instructional resource rather than an autonomous teaching agent. Full article

Knowledge Tracing, which estimates how students’ knowledge evolves during interactions with educational content, is a cornerstone of Intelligent Tutoring Systems. While deep learning models achieve superior predictive performance in this task, they lack interpretability, a limitation that is particularly critical in educational contexts. We introduce gTransformer, a new type of grounded Transformer model bridging deep learning performance with intrinsic interpretability through representational grounding. It adds theory-based parameters to input interaction sequences and uses attention mechanisms to transform them into latent representations. These are projected into enriched parameters that incorporate historical learning context while preserving semantics. Validation demonstrates: (1) structural encoding around theoretical concepts (probing selectivity $\Delta R^2>0.5$); (2) semantic alignment; and (3) functional alignment with quantified confidence. Results show that gTransformer achieves predictive performance competitive with state-of-the-art architectures while offering intrinsically interpretable predictions. The trade-off is characterised by a significant Area Under the Curve (AUC) gain over traditional theory-based models (+19.9%), with a minimal cost (3.9%) relative to non-interpretable configurations. Critically, gTransformer enables context-aware personalisation by differentiating students based on longitudinal learning trajectories rather than immediate responses, capturing patterns that traditional models cannot represent. This offers a practical path toward adaptive instruction driven by artificial intelligence that is both accurate and interpretable. Full article

Large Language Models (LLMs) demonstrate remarkable performance across a variety of natural language understanding tasks, yet their sensitivity to emotional framing in user prompts remains underexplored. This paper presents an empirical study investigating how four emotional tones—joy, apathy, anger, and fear—affect LLM performance on the SuperGLUE benchmark. We evaluate five instruction-tuned, open-weight models across eight diverse tasks, systematically modulating input prompts with affective cues while keeping semantic content constant. Results reveal that prompts framed with joy and apathy lead to consistently higher accuracy, with gains of up to 4.5 percentage points compared to fear-framed inputs, which yield the lowest performance. These findings demonstrate that affective modulation in user prompts measurably impacts LLM reasoning and task outcomes, suggesting that emotional framing is not merely stylistic but functionally relevant to model behavior. Our study provides a reproducible experimental framework and an open-source prompt set, offering a foundation for future research on affect-aware prompting strategies and their implications in human–AI interaction. Full article

Generative artificial intelligence and foundation models have changed machine learning by allowing systems to produce readable text, realistic images, and other multimodal content with little direct input from a user. Foundation models are large neural networks trained on very large and varied datasets, and they form the core of many current generative AI (GenAI) systems. Their rapid development has led to major advances in areas like natural language processing, computer vision, multimodal learning, and robotics. Examples include GPT, LLaMA, and diffusion-based architectures, such as models often used for image generation. Systems such as Stable Diffusion show this shift by illustrating how AI can interpret information, draw basic inferences, and produce new outputs using more than one type of data. This review surveys common foundation model architectures and examines what they can do in generative tasks. It reviews Transformer, diffusion, and multimodal architectures, focusing on methods that support scaling and transfer across domains. The paper also reviews key approaches to pretraining and fine-tuning, including self-supervised learning, instruction tuning, and parameter-efficient adaptation, which support these systems’ ability to generalize across tasks. In addition to the technical details, this review discusses how GenAI is being used for text generation, image synthesis, robotics, and biomedical research. The study also notes continuing challenges, such as the high computing and energy demands of large models, ethical concerns about data bias and misinformation, and worries about privacy, reliability, and responsible use of AI in real settings. This review brings together ideas about model design, training methods, and social implications to point future research toward GenAI systems that are efficient, easy to interpret, and reliable, while supporting scientific progress and ethical responsibility. Full article

Heritage speakers’ language acquisition is a complex process that is affected by linguistic, social, cultural, and affective factors. Studies on heritage speakers (HSs) have primarily focused on challenges HSs face in the classroom and scarcely investigated these challenges outside of instructional settings. This study addresses this gap by exploring the lived experiences of two young adult Spanish HSs outside of educational settings through a series of interviews to create personal narratives of their HL and experiences. Through Narrative Interpretative Phenomenological Analysis (NIPA), three main themes emerged from these narratives: (1) Spanish heritage language (HL) knowledge and language use, (2) emotional factors that hinder language knowledge and language use, and (3) self-positioning towards SHL and culture. The findings indicated that the participants’ experiences with their Spanish heritage language (SHL) were profoundly impacted by the nature of language input they received, hostile environments, and negative interactions with members of their communities, which led to emotional distress and communicative avoidance. This situated study also offers potential conceptual and community-based implications for the Spanish HSs. Full article

This study investigates how integrating the 4PADAFE instructional design methodology with generative artificial intelligence (GAI) tools helps develop innovative, pedagogically sound digital learning environments in higher education. To meet the demand for scalable and flexible instructional models, 4PADAFE offers a seven-phase, iterative framework that connects pedagogical goals with the creative use of AI-powered tools. Using a qualitative exploratory approach, 20 Systems Engineering students applied the methodology to collaboratively create a four-week Massive Open Online Course (MOOC) titled “Generative Artificial Intelligence Tools for University Teaching.” They utilized ChatGPT, DALL·E, and Gamma to produce educational materials without direct input from subject-matter experts. Data collection included semi-structured interviews, non-participant observation, and analysis of student-created artifacts. The findings revealed increased learner autonomy, creativity, and digital skills, along with more efficient instructional design processes supported by prompt engineering and real-time feedback. The structured 4PADAFE framework helped participants align AI-generated content with specific learning outcomes while maintaining ethical safeguards. This study concludes that, with proper guidance and a systematic framework, students with technical backgrounds can serve as effective instructional designers, demonstrating the potential of combining structured methodologies and GAI to democratize high-quality course development in digital higher education. Full article

This study investigates the pedagogical value of integrating AI-supported feedback with Error Analysis in university-level English as a Foreign Language (EFL) writing instruction, where English is the target language (TL). Adopting a comparative, corpus-based design, the research examines whether AI-mediated feedback can complement traditional teacher-led Error Analysis in reducing recurrent errors, improving grammatical accuracy, and supporting revision practices among Spanish L1 learners of English at the B2 (CEFR) level. Seventy participants completed two writing tasks over a twelve-week period, generating a learner corpus that was randomly assigned to two groups: AI-assisted feedback and teacher-mediated feedback. Quantitative Error Analysis and learner-perception surveys were conducted to assess both linguistic outcomes and attitudinal responses. Results indicate that students receiving AI-assisted feedback demonstrated lower rates of error repetition (25%) compared to those receiving teacher-based correction (40%), particularly in subject–verb agreement, preposition use, tense selection, and L1-induced lexical transfer in L2 English writing. Survey findings further reveal higher perceived levels of clarity, usefulness, and immediacy for AI-generated feedback, although participants continued to value teacher input for higher-order writing concerns. Overall, the findings suggest that AI-supported Error Analysis can contribute to short-term error reduction and foster learner autonomy. This study highlights the potential of blended and mixed feedback models within a focused pedagogical context and underscores the need for longitudinal research examining long-term retention, pragmatic development, and cross-context generalizability. Full article

A safe physical human–robot interaction (pHRI) in rehabilitation requires reliable perception and low-latency decision making under heterogeneous and unreliable sensor inputs. This paper presents a multimodal sensor-fusion-based safety framework that integrates physical state estimation, semantic information fusion, and an edge-deployed large language model (LLM) for real-time pHRI safety control. A dynamics-based virtual sensing method is introduced to estimate internal joint torques from external force–torque measurements, achieving a normalized mean absolute error of $18.5\%$ in real-world experiments. An asynchronous semantic state pool with a time-to-live mechanism is designed to fuse visual, force, posture, and human semantic cues while maintaining robustness to sensor delays and dropouts. Based on structured multimodal tokens, an instruction-tuned edge LLM outputs discrete safety decisions that are further mapped to continuous compliant control parameters. The framework is trained using a hybrid dataset consisting of limited real-world samples and LLM-augmented synthetic data, and evaluated on unseen real and mixed-condition scenarios. Experimental results show reliable detection of safety-critical events with a low emergency misdetection rate, while maintaining an end-to-end decision latency of approximately 223 ms on edge hardware. Real-world experiments on a rehabilitation robot demonstrate effective responses to impacts, user instability, and visual occlusions, indicating the practical applicability of the proposed approach for real-time pHRI safety monitoring. Full article

An empathetic chatbot is an essential component of intelligent mental healthcare. We adopted implicitly empathy prompting (IEP) by decomposing empathy into supportive dialogue, paraphrased response, emotional understanding, and attitude expression, referred to as the four features of empathy decomposition. IEP is based on lightweight language multi-agents (LLM-Agents) to generate empathy dialogue. The approach contrasts with the explicitly defined empathy of simply prompting a model to be empathetic. Three datasets for the four features scenario were generated by using the Generative Pre-Trained Transformer (GPT)-4o model, with cases in finance, family, and health issues. For each dataset, 30 examples were randomly selected and examined as input prompting onto six lightweight language models. These models include Mistral (7B), Phi-4 (14B), StableLM2 (12B), Tulu3 (8B), Neural-chat (7B), and Llama 3.1-Instruct (8B). After that, the output was evaluated by using GPT-4o to calculate empathy perception scores (EP scores). The average EP scores on three datasets for implicit/explicit empathy prompting ranged from 1 to 10. The final evaluation results are as follows: (1) implicitly empathy prompting (IEP): Mistral (8.83), Phi-4 (8.96), StableLM2 (9.03), Tulu3 (7.24), Neural-chat: (8.03), Llama 3.1-Instruct (8.74); (2) Explicitly Empathy Prompting (EEP): Mistral (7.52), Phi-4 (8.55), StableLM2 (7.78), Tulu3 (7.67), Neural-chat (8.35), Llama 3.1-Instruct (8.76). Among these values, three models (Mistral, Phi-4, and StableLM2) achieve higher and stable EP scores obviously. The other models (Tulu3, Neural-chat, and Llama 3.1-Instruct) keep comparable EP scores. Our experiment findings showed that the prompt engineering method with the IEP approach could significantly outperform EEP. Full article

In Tanzanian secondary education, Swahili-language-based question-answering systems currently face systemic disparities and linguistic barriers, which undermine the fairness and justice of the educational system. While Large Language Models (LLMs) offer scalable instructional support, they typically lack curriculum grounding, which causes them to perform unreliably in low-resource languages. This study introduces a Curriculum-Aware Retrieval-Augmented Generation (RAG) framework designed to be a linguistically inclusive AI tutor. The architecture combines hybrid dense–lexical retrieval, cross-encoder reranking, and metadata-based curriculum alignment to ensure factual, grade-appropriate responses. We evaluate five distinct generative models using a stratified 500-question Golden Dataset covering English, Swahili, and code-switched inputs. Findings indicate that there is a significant trade-off between scale and deployability. Although high-capacity LLMs provide useful reference performance, Qwen2.5-0.5B offers the most realistic trade-off between quality and deployability in low-resource settings. Under the proposed curriculum-aware pipeline, Qwen2.5-0.5B attains the best answer quality (F1: 32.7%), achieves strong grounding faithfulness (83.0%, validated by human evaluation), and maintains low end-to-end latency suitable for interactive classroom use (≤1.24 s). Notably, considering the limited size of the code-switched evaluation subset, our framework demonstrates promising capabilities in handling Swahili–English code-switched inputs, narrowing the observed performance gap between Swahili and English through improved semantic accuracy. These results provide initial empirical evidence that curriculum-aligned RAG can enable Small Language Models (SLMs) to serve as quality, safe, and sustainable educational assistants in low-resource Global South contexts. Full article

Adverse weather removal aims to restore images degraded by haze, rain, or snow. However, existing unified models often rely on implicit degradation cues, making them vulnerable to inaccurate weather perception and insufficient semantic guidance, which leads to over-smoothing or residual artifacts in real scenes. In this work, we propose AWR-VIP, a prior-guided adverse weather removal framework that explicitly extracts semantic and perceptual priors using a frozen vision–language model (VLM). Given a degraded input, we first employ a degradation-aware prompt extractor to produce a compact set of semantic tags describing key objects and regions, and simultaneously perform weather-type perception by prompting the VLM with explicit weather definitions. Conditioned on the predicted weather type and selected tags, the VLM further generates two levels of restoration guidance: a global instruction that summarizes image-level enhancement goals (e.g., visibility/contrast) and local instructions that specify tag-aware refinement cues (e.g., recover textures for specific regions). These textual outputs are encoded by a text encoder into a pair of priors (${\mathbf{P}}_{global}$ and ${\mathbf{P}}_{local}$), which are injected into a UNet-based restorer through global-prior-modulated normalization and instruction-guided attention, enabling weather-adaptive and content-aware restoration. Extensive experiments on a combined benchmark show that AWR-VIP consistently outperforms state-of-the-art methods. Moreover, the VLM-derived priors are plug-and-play and can be integrated into other restoration backbones to further improve performance. Full article

Dozens of studies published on processing instruction (PI) make it one of the highly investigated areas within the field of second language acquisition (SLA). This review article provides an overview of 30 years of research in this area. The article first provides a brief account of the theoretical underpinnings of PI, its features, and the guidelines for devising structured input (SI) activities, which are the main causative factor accounting for the effects of PI. It then proceeds with a review of empirical studies carried out across six strands of research on PI and offers a detailed synthesis of the key findings emerging from each strand. The article ends with a general conclusion based on the findings of research to date and outlines directions for future research. Full article

The accurate interpretation of land cover changes in multi-temporal satellite imagery is critical for Earth observation. However, existing methods typically yield static outputs—such as binary masks or fixed captions—lacking interactivity and user guidance. To address this limitation, we introduce remote sensing image change analysis (RSICA), a novel paradigm that enables the instruction-guided, multi-turn exploration of temporal differences in bi-temporal images through visual question answering. To realize RSICA, we propose DeltaVLM, a vision language model specifically designed for interactive change understanding. DeltaVLM comprises three key components: (1) a fine-tuned bi-temporal vision encoder that independently extracts semantic features from each image in the input pair; (2) a visual difference perception module with a cross-semantic relation measuring (CSRM) mechanism to interpret changes; and (3) an instruction-guided Q-former that selects query-relevant change features and aligns them with a frozen large language model to generate context-aware responses. We also present ChangeChat-105k, a large-scale instruction-following dataset containing over 105k diverse samples. Extensive experiments show that DeltaVLM achieves state-of-the-art performance in both single-turn captioning and multi-turn interactive change analysis, surpassing both general multimodal models and specialized remote sensing vision language models. Full article

The rapid adoption of large language models (LLMs) in corporate and governmental systems has raised critical security concerns, particularly prompt injection attacks exploiting LLMs’ inability to differentiate control instructions from untrusted user inputs. This study systematically benchmarks neural network architectures for malicious prompt detection, emphasizing robustness against character-level adversarial perturbations—an aspect that remains comparatively underemphasized in the specific context of prompt-injection detection despite its established significance in general adversarial NLP. Using the Malicious Prompt Detection Dataset (MPDD) containing 39,234 labeled instances, eight architectures—Dense DNN, CNN, BiLSTM, BiGRU, Transformer, ResNet, and character-level variants of CNN and BiLSTM—were evaluated based on standard performance metrics (accuracy, F1-score, and AUC-ROC), adversarial robustness coefficients against spacing and homoglyph perturbations, and inference latency. Results indicate that the word-level 3_Word_BiLSTM achieved the highest performance on clean samples (accuracy = 0.9681, F1 = 0.9681), whereas the Transformer exhibited lower accuracy (0.9190) and significant vulnerability to spacing attacks (adversarial robustness ${\mathsf{\rho }}_{spacing}=0.61$). Conversely, the Character-level BiLSTM demonstrated superior resilience (${\mathsf{\rho }}_{spacing}=1.0$, ${\mathsf{\rho }}_{homoglyph}=0.98$), maintaining high accuracy (0.9599) and generalization on external datasets with only 2–4% performance decay. These findings highlight that character-level representations provide intrinsic robustness against obfuscation attacks, suggesting Char_BiLSTM as a reliable component in defense-in-depth strategies for LLM-integrated systems. Full article

Large Language Models (LLMs) are vulnerable to Indirect Prompt Injection Attacks (IPIAs), where malicious instructions are embedded within external content rather than direct user input. This study presents an embedding-based detection approach that analyses the semantic relationship between user intent and external content, enabling the early identification of IPIAs that conventional defences overlook. We also provide a dataset of 70,000 samples, constructed using 35,000 malicious instances from the Benchmark for Indirect Prompt Injection Attacks (BIPIA) and 35,000 benign instances generated using ChatGPT-4o-mini. Furthermore, we performed a comparative analysis of three embedding models, namely OpenAI text-embedding-3-small, GTE-large, and MiniLM-L6-v2, evaluated in combination with XGBoost, LightGBM, and Random Forest classifiers. The best-performing configuration using OpenAI embeddings with XGBoost achieved an accuracy of 97.7% and an F1-score of 0.977, matching or exceeding the performance of existing IPIA detection methods while offering practical deployment advantages. Unlike prevention-focused approaches that require modifications to the underlying LLM architecture, the proposed method operates as a model-agnostic external detection layer with an average inference time of 0.001 ms per sample. This detection-based approach complements existing prevention mechanisms by providing a lightweight, scalable solution that can be integrated into LLM pipelines without requiring architectural changes. Full article