Large Language Models (LLMs) are trained on web-scale corpora, increasing the risk that benchmark test data appears in training sets and inflates reported performance. We present a systematic literature review of 55 studies on LLM benchmark contamination through late 2025. Our contributions are: (1) a four-tier contamination taxonomy (Exact, Syntactic, Semantic, Task-Level; T1–T4); (2) a comparative analysis of five detection families (string-matching, likelihood-based, membership inference, LLM-prompted detection, and benchmark auditing), including access assumptions and failure modes; (3) a synthesis of contamination evidence on MMLU, GSM8K, HumanEval, and HellaSwag by measurement construct; (4) a comparative evaluation of mitigation strategies across lifecycle points, access assumptions, and evidence maturity; and (5) a Contamination Transparency Card (CTC) framework for future releases. Across studies, no detection method is consistently reliable across contamination tiers, model-access settings, and training stages. We identify instruction tuning as a persistent blind spot, note that RL/post-training contamination auditing is only beginning to mature, and report inflation estimates spanning roughly 6%–40% under benchmark- and setting-dependent assumptions.

Medical question answering is a high-stakes setting where factual errors can have serious consequences. Retrieval-augmented generation (RAG) is widely viewed as a promising solution, and prior work has reported substantial gains for large medical QA models. We revisit this assumption across a broad range of open-weight instruction-tuned models spanning 7B to 72B parameters. Across five models, ten biomedical QA datasets, four retrieval methods, and four retrieval corpora, we find that retrieval yields only small and inconsistent improvements over a no-retrieval baseline, typically within 1-2 points. In contrast, the choice of backbone model has a much larger effect than the choice of retriever or corpus, and expert and layman retrieval sources perform similarly in most settings. These results suggest that the main bottleneck is not retrieval quality alone, but the model's limited ability to use retrieved evidence effectively.

Real-time cognitive load assessment from eye-tracking signals could potentially enable adaptive human-centered-AI such as safety-critical applications such as driver vigilance monitoring or automated flight deck assistance, yet two challenges persist: handling frequent data missingness from blinks and tracking failures, and efficiently modeling long-range temporal dependencies. We propose MambaGaze, a framework that addresses these challenges through 1) XMD encoding, which augments raw features with observation masks and time-deltas to explicitly model data uncertainty, and 2) bidirectional Mamba-2, which captures temporal dependencies with linear computational complexity. Experiments on CLARE and CL-Drive datasets under leave-one-subject-out evaluation show that MambaGaze achieves 76.8% and 73.1% accuracy, respectively, outperforming CNN, Transformer, ResNet, and VGG baselines by 4-12 percentage points. Edge deployment benchmarks on NVIDIA Jetson platforms demonstrate real-time inference at 43-68 FPS with power consumption below 7.5W, confirming feasibility for wearable cognitive load monitoring.

Real-time cognitive load assessment is essential for adaptive human-computer interaction but remains challenging due to limited labeled data and poor cross-subject generalization. Recent ECG foundation models pre-trained on millions of clinical recordings offer rich representations, but cannot be directly applied to wearable devices due to sensor configuration mismatch and task differences. In this paper, we propose CogAdapt, a framework that adapts clinical ECG foundation models to wearable cognitive load assessment. CogAdapt introduces LeadBridge, a learnable adapter that transforms 3-lead wearable signals into anatomically consistent 12-lead representations, and ProFine, a progressive fine-tuning strategy that gradually unfreezes encoder layers while preventing catastrophic forgetting. Evaluations on two public datasets (CLARE and CL-Drive) under leave-one-subject-out cross-validation show that CogAdapt substantially outperforms baselines trained from scratch, achieving macro-F1 scores of 0.626 and 0.768. These results demonstrate the promise of foundation model adaptation for subject-independent cognitive load assessment from wearable sensors.

Prompt-based evaluations suggest that large language models (LLMs) perform poorly on time series classification, raising doubts about whether they encode meaningful temporal structure. We show that this conclusion reflects limitations of prompt-based generation rather than the model's representational capacity by directly comparing prompt outputs with linear probes over the same internal representations. While zero-shot prompting performs near chance, linear probes improve average F1 from 0.15-0.26 to 0.61-0.67, often matching or exceeding specialized time series models. Layer-wise analyses further show that class-discriminative time series information emerges in early transformer layers and is amplified by visual and multimodal inputs. Together, these results demonstrate a systematic mismatch between what LLMs internally represent and what prompt-based evaluation reveals, leading current evaluations to underestimate their time series understanding.

The rise of Internet of Things (IoT) devices has led to the development of numerous time-sensitive applications that require quick responses and low latency. Fog computing has emerged as a solution for processing these IoT applications, but it faces challenges such as resource allocation and job scheduling. Therefore, it is crucial to determine how to assign and schedule tasks on Fog nodes. This work aims to schedule tasks in IoT while minimizing the total energy consumption of nodes and enhancing the Quality of Service (QoS) requirements of IoT tasks, taking into account task deadlines. This paper classifies Fog nodes into two categories based on their traffic level: low and high. It schedules short-deadline tasks on low-traffic nodes using an Improved Golden Eagle Optimization (IGEO) algorithm, an enhancement that utilizes genetic operators for discretization. Long-deadline tasks are processed on high-traffic nodes using reinforcement learning (RL). This combined approach is called the Reinforcement Improved Golden Eagle Optimization (RIGEO) algorithm. Experimental results demonstrate that RIGEO achieves up to a 29% reduction in energy consumption, up to an 86% improvement in response time, and up to a 19% reduction in deadline violations compared to state-of-the-art algorithms.

The rapid integration of generative artificial intelligence into education has driven digital transformation in e-teaching, yet user perceptions of AI educational apps remain underexplored. This study performs a sentiment-driven evaluation of user reviews from top AI ed-apps on the Google Play Store to assess efficacy, challenges, and pedagogical implications. Our pipeline involved scraping app data and reviews, RoBERTa for binary sentiment classification, GPT-4o for key point extraction, and GPT-5 for synthesizing top positive/negative themes. Apps were categorized into seven types (e.g., homework helpers, math solvers, language tools), with overlaps reflecting multifunctional designs. Results indicate predominantly positive sentiments, with homework apps like Edu AI (95.9% positive) and Answer.AI (92.7%) leading in accuracy, speed, and personalization, while language/LMS apps (e.g., Teacher AI at 21.8% positive) lag due to instability and limited features. Positives emphasize efficiency in brainstorming, problem-solving, and engagement; negatives center on paywalls, inaccuracies, ads, and glitches. Trends show that homework helpers outperform specialized tools, highlighting AI's democratizing potential amid risks of dependency and inequity. The discussion proposes future ecosystems with hybrid AI-human models, VR/AR for immersive learning, and a roadmap for developers (adaptive personalization) and policymakers (monetization regulation for inclusivity). This underscores generative AI's role in advancing e-teaching by enabling ethical refinements that foster equitable, innovative environments.

Age-related macular degeneration (AMD) and choroidal neovascularization (CNV)-related conditions are leading causes of vision loss worldwide, with optical coherence tomography (OCT) serving as a cornerstone for early detection and management. However, deploying state-of-the-art deep learning models like ConvNeXtV2-Large in clinical settings is hindered by their computational demands. Therefore, it is desirable to develop efficient models that maintain high diagnostic performance while enabling real-time deployment. In this study, a novel knowledge distillation framework, termed KD-OCT, is proposed to compress a high-performance ConvNeXtV2-Large teacher model, enhanced with advanced augmentations, stochastic weight averaging, and focal loss, into a lightweight EfficientNet-B2 student for classifying normal, drusen, and CNV cases. KD-OCT employs real-time distillation with a combined loss balancing soft teacher knowledge transfer and hard ground-truth supervision. The effectiveness of the proposed method is evaluated on the Noor Eye Hospital (NEH) dataset using patient-level cross-validation. Experimental results demonstrate that KD-OCT outperforms comparable multi-scale or feature-fusion OCT classifiers in efficiency-accuracy balance, achieving near-teacher performance with substantial reductions in model size and inference time. Despite the compression, the student model exceeds most existing frameworks, facilitating edge deployment for AMD screening.

The rapid integration of generative artificial intelligence (AI) in higher education since 2023 has outpaced institutional preparedness, creating a persistent gap between student practices and established ethical standards. This paper draws on mixed-method surveys and a focused literature review to examine student motivations, ethical dilemmas, gendered responses, and institutional readiness for AI adoption. We find that 92% of students use AI tools primarily to save time and improve work quality, yet only 36% receive formal guidance, producing a de facto "shadow pedagogy" of unguided workflows. Notably, 18% of students reported integrating AI-constructed material into assignments, which suggests confusion about integrity expectations and compromises the integrity of the assessment. Female students expressed greater concern about abuse and distortion of information than male students, revealing a gendered difference in awareness of risk and AI literacies. Correspondingly, 72% of educators use AI, but only 14% feel at ease doing so, reflecting limited training and uneven policy responses. We argue that institutions must adopt comprehensive AI literacy programs that integrate technical skills and ethical reasoning, alongside clear AI-use policies and assessment practices that promote transparency. The paper proposes an Ethical AI Integration Model centered on literacy, gender-inclusive support, and assessment redesign to guide responsible adoption, protect academic integrity, and foster equitable educational outcomes in an AI-driven landscape.

Text summarization is a valuable method for extracting important details from large volumes of text data, facilitating tasks like text data analysis. Various text summarization techniques have been developed over time, with some focusing on selecting and summarizing short sentences, while others overlook the semantic relationship between sentences. Extractive document summarization involves learning cross-sentence relations, a critical aspect that has been extensively explored using various approaches. One effective method is to employ neural networks based on graphs, which offer an intricate structure capable of obtaining relations among sentences. In this paper, we present a contextualized heterogeneous graph neural network for extractive text summarization (ConHGNN-SUM), incorporating semantic nodes that extend beyond individual sentences, and emphasizes the importance of capturing the relationship between selected sentences as a final step in the summarization process. These extra nodes function as intermediaries connecting sentences and enhancing the interrelationships between them. Our model enhances conventional graph-based extractive methods and delivers comparable performance to other advanced systems for extractive summarization.