2833 papers
This survey presents a unified representation learning framework that traces the convergence of discriminative, multimodal, and generative paradigms in modern computer vision—from CNNs and Vision Transformers to foundation and diffusion models—arguing for a cohesive perspective to advance general-purpose visual intelligence.
This study reports that the WordBinary AI-text detection model achieved perfect classification accuracy (100%) on a specific benchmark of 11,000 samples comprising 9,000 AI-generated texts and 2,000 pre-2015 human academic papers, while acknowledging that these results require further independent validation and broader testing before generalizing to universal accuracy.
This paper introduces and validates a systematic framework for auditing the reliability of Bengali LLM benchmarks by detecting cross-lingual data contamination and measuring robustness against perturbations, revealing that initial pilot results show performance barely above chance levels.
This review examines the transformative impact of AI-driven and conventional 3D image reconstruction techniques in medical imaging, highlighting how deep learning, augmented reality, and optimized algorithms enhance diagnostic precision and surgical planning while addressing ongoing challenges like computational demands and noise.
This study demonstrates that JuriBERT, a French legal domain-adapted model, most accurately predicts the outcomes of over 157,000 French administrative tribunal decisions based solely on factual and procedural backgrounds, achieving a macro F1 score of 0.718 and highlighting the predictive signal embedded in case facts prior to adjudication.
This paper presents a controlled evaluation demonstrating that repository-scoped knowledge graph context selection effectively generates ontology validation tests for the multi-module TKTOnto artifact by automatically isolating relevant modules, satisfying a deterministic relevance contract, and reducing input tokens by 2.24%.
This study proposes a monotone, redundancy-aware decision-burden method for explainable human oversight in human–robot collaboration that ensures safety conditions are never overridden by aggregate scores, demonstrating improved predictive performance and mathematical monotonicity over unadjusted models in extensive computational simulations while noting the need for future field validation.
This paper proposes an automated hyperparameter optimization framework for transfer learning-based machine fault diagnosis using a hybrid Genetic–Simulated Annealing algorithm, which demonstrates robust performance improvements over baseline methods across diverse industrial scenarios involving data scarcity and domain shifts.
This paper argues that authorship attribution is structurally impossible for convergent assessment formats like multiple-choice questions because correct responses contain zero residual entropy, rendering current AI-integrity policies based on essay detection unsound for quantitative disciplines and necessitating modality-specific controls.
This study demonstrates that the PEFM-CMAE model, which integrates acoustic-prosodic features and explicit text-audio incongruity, significantly outperforms text-only baselines in detecting spoken sarcasm on the MUStARD corpus, though its performance drops when generalizing to unseen shows.