4158 papers
This paper proposes SS-MAPPO, an improved Multi-Agent Proximal Policy Optimization framework incorporating top-K sparse attention and Gaussian latent policies, to solve the energy-efficient, decentralized task offloading problem in dynamic multi-UAV edge computing environments modeled as a Dec-POMDP.
This paper evaluates and compares the performance of classical geometric rules, Random Forest, and deep learning architectures (PointNet and PointNet++) in classifying synthetic 3D particle shapes, revealing that while Random Forest achieves perfect accuracy on standard and geometry-based tests, PointNet++ with specific configurations yields the best results on challenging boundary cases.
While supervised foundation-model probes outperform generative reconstruction methods for artifact detection in digital pathology under consistent scanning conditions, they suffer catastrophic failure when magnification shifts occur, necessitating verification of acquisition parameters and the use of reconstruction-based safeguards to ensure robust quality control.
This paper proposes a truth- and intent-agnostic ontology for analyzing communicative influence at scale, which structures messages into four orthogonal dimensions (Method, Narrative, Polarization, and AboutObject) grounded in observable cues to support both close reading and large-scale computational analysis of strategic communication beyond traditional disinformation frameworks.
This study demonstrates that a 17-layer residual convolutional neural network (DnCNN) significantly outperforms classical Gaussian, median, and non-local means filters in denoising synthetic industrial CT images, achieving superior peak signal-to-noise ratio and structural similarity across all tested dose levels while highlighting the need for future validation on real-world scanner data.
The paper proposes **Cranio-Diff**, a diffusion-based framework that integrates skull-conditioned structural guidance and biometric text prompts to reconstruct faces from 2D X-ray skull images, demonstrating superior performance in image quality and retrieval accuracy on a novel, large-scale synthetic dataset designed for forensic identification.
This paper proposes a Dual-Gated Dueling Double Deep Q-Networks framework for offline reinforcement learning in sepsis resuscitation, which effectively mitigates Q-value overestimation and promotes safer, more conservative fluid and vasopressor dosing compared to standard models and current clinical practice.
This paper quantifies the implementation-specific timing leakage of 16-byte early-exit secret comparisons versus fixed-work alternatives on ESP32 and ESP8266 microcontrollers, demonstrating that matching-prefix length significantly increases execution time on these platforms while fixed-work routines maintain constant timing, though the study does not confirm binary-level constant-time behavior or remote exploitability.
ProtoSPM is a protocol-structured, parameter-efficient pretraining framework that constructs a fixed, reusable traffic encoder through masked reconstruction of bidirectional flow matrices, enabling high-performance encrypted traffic classification across diverse tasks without updating the backbone weights.
This study proposes a CBAM-enhanced MaxViT-Tiny model for 8-class gastrointestinal endoscopy image classification that achieves high accuracy and explainability on clean data but demonstrates significant vulnerability to adversarial attacks and poor cross-attack generalization, highlighting the critical need for adversarial training and multi-center validation before clinical deployment.