2787 papers
The paper introduces PHCTQA, a lightweight and rapid AI system that utilizes a physics-informed multi-view and multi-instance learning approach to accurately identify ten head-thorax NCCT quality defects in real-time, significantly outperforming junior radiographers and reducing rescanning rates in resource-constrained settings.
This paper proposes SA-RGLP, a novel framework that enhances metaphor detection by integrating syntax-aware incongruity modeling with LLM-derived soft priors through residual graph label propagation to capture global semantic structures and cross-sample relational reasoning, achieving state-of-the-art results on multiple benchmarks.
This paper introduces a reward-action validity audit to demonstrate that a Deep Q-Network notebook claiming to learn English morphology actually fails to acquire grammatical rules due to fundamental flaws in its reward function, action expressiveness, state encoding, and evaluation methodology.
This empirical study demonstrates that in hand sign image classification, increasing CNN capacity does not guarantee better performance and that dropout regularization, particularly at architecture-dependent rates, significantly outperforms L1 and L2 penalties, with a medium-sized model achieving the highest accuracy of 96.66%.
This paper introduces LMS-FAIR-India, a fully synthetic, FAIR-compliant dataset of 1.7 million LMS interaction events from a private Indian university that enables privacy-preserving learning analytics research without using real student data.
This paper empirically evaluates the coverage and complementarity of three dominant agentic AI risk taxonomies (OWASP-ASI, MSFT-AIRT, and NIST AI RMF) across 131 real-world incidents to provide evidence-based guidance for practitioners and identify specific gaps for future framework revisions.
This paper proposes a novel memetic ant colony system algorithm integrated with variable neighborhood descent to solve the multiple machine scheduling problem in additive manufacturing by simultaneously optimizing energy costs, makespan, and part orientations, demonstrating significant efficiency improvements over existing methods.
This paper introduces the Curvature-aware Information Bottleneck (CurvIB) framework, a theoretically grounded model compression technique based on information geometry and the curvature-information duality theorem, which unifies curvature-sensitive adaptive pruning, Wasserstein-aware optimal quantization, and optimal transport-based accuracy recovery to significantly enhance deep learning model performance under extreme resource constraints.
This study demonstrates that on the BUSI dataset, the performance gains from various lightweight attention mechanisms in breast ultrasound lesion segmentation are statistically negligible and below the resolution of the data, whereas factors like network depth and data leakage (via checkpoint selection or near-duplicates) produce significantly larger, resolvable effects.
This paper presents a fully reproducible NLP pipeline, including a curated corpus, a six-category ESG taxonomy, and a fine-tuned Spanish BERT classifier, designed to quantify the divergence between corporate ESG rhetoric and observable outcomes in media coverage through a novel "Say-Do Gap Index" while addressing construct validity and cross-context applicability.