3020 papers
This study introduces CodePori, a large-scale multi-agent system for autonomous software development, and empirically evaluates its real-world applicability through participant feedback to identify key challenges and opportunities beyond standard benchmark metrics.
This study proposes and evaluates strategies for redesigning Computer Science homework to leverage, rather than prohibit, AI tools, demonstrating that iterative learning cycles, strategic timing, and continuous feedback can enhance genuine student understanding while maintaining academic integrity.
This paper introduces Res6-Deep11, a novel unsupervised image reconstruction model that effectively detects and segments oil spills in SAR images without requiring pixel-level ground truth labels, achieving superior performance over state-of-the-art methods with an F1-score of 77.33%.
This paper introduces DTap, the first controllable and interactive red-teaming platform for AI agents, which includes an autonomous red-teaming agent (DTap-Red) and a large-scale benchmark (DTap-Bench) to systematically evaluate and expose security vulnerabilities across 14 real-world domains.
This paper presents a scalable framework for operationalizing trustworthy AI in healthcare by utilizing an Error Analysis Decision Tree and novel methodological innovations to uncover hidden subgroup failures and mitigate bias, thereby translating abstract fairness principles into verifiable engineering objectives that ensure compliance with regulations like the EU AI Act.
This systematic review of 154 studies (2021–2025) reveals that while AI for IoT/IIoT security is heavily focused on intrusion detection, current research lacks operational validity due to scarce labels and resource constraints, necessitating a shift toward standardized, drift-aware, and explainable deployment protocols.
This paper introduces the constraint-aware Decomposition–Recomposition Process Element Annotation (DR-PEA) method to mitigate Process Semantic Collapse in large language models by transforming unstructured process text into standardized structured data through a three-stage pipeline of semantic decomposition, process recomposition, and constraint mapping, achieving superior performance over existing baselines.
This study compares the classification accuracy and explainability stability of DenseNet121 and Vision Transformer models on breast histopathology images, revealing a critical trade-off where DenseNet121 achieves higher accuracy and localized explanations while the Vision Transformer demonstrates greater robustness of explanations under input perturbations.
This study presents a three-phase deep learning framework that combines LSTM-based pH prediction with Deep Reinforcement Learning control, validated against African soil profiles using synthetic data, to provide a scalable and transferable solution for optimizing mine reclamation in data-scarce Sub-Saharan regions.
This study demonstrates that Deep Set-based aggregation leveraging self-attention outperforms traditional methods in modeling variable-length, high-frequency wearable physical activity data, while suggesting that simpler aggregation strategies remain sufficient for more stable, low-frequency measurements.