Projected Gradient Descent
PulseAugur coverage of Projected Gradient Descent — every cluster mentioning Projected Gradient Descent across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New research reveals potent backdoor attack methods targeting LLM agents
Two new research papers explore vulnerabilities in large language model (LLM) agents, focusing on backdoor attacks. The first paper, AGENTQ, introduces a method to create attacks that are effective even after quantizati…
-
Adversarial training for P2P lending models shows mixed robustness across attack types
Researchers have evaluated the robustness of machine learning models used in peer-to-peer lending against various adversarial attacks. The study found that while adversarial training significantly improves a model's def…
-
New AI attack method exploits text-guided diffusion for chest X-ray vulnerabilities
Researchers have developed a novel text-guided diffusion-based adversarial framework to test the vulnerability of AI models used in chest X-ray (CXR) interpretation. Unlike traditional pixel-space attacks, this method u…
-
New benchmark evaluates Kolmogorov-Arnold Network robustness against adversarial attacks
Researchers have developed KAN-Robust-Bench, a new benchmark designed to evaluate the robustness of Kolmogorov-Arnold Networks (KANs) against adversarial evasion attacks. The study explores both certified and empirical …
-
New CIVA attack method targets visual world-model agents
Researchers have developed a new method called Critic-Induced Value-Subspace Attacks (CIVA) to target visual world-model agents. These agents, like DreamerV3, operate using a recurrent latent state, making them resilien…
-
New study questions PGD's ability to measure AI model robustness
A new study published on arXiv investigates the effectiveness of Projected Gradient Descent (PGD) in evaluating adversarial robustness for convolutional neural networks. Researchers found that while PGD is commonly used…
-
New theory and algorithms advance coherent imaging techniques
Researchers have developed a new theoretical framework and algorithms for multilook coherent imaging, a technique used in applications like digital holography and synthetic aperture radar. The paper provides the first t…
-
New AI tool PPAPlace optimizes chip placement for better performance
Researchers have developed PPAPlace, a novel AI-driven system designed to optimize chip placement for improved performance, power, and area (PPA). Unlike traditional methods that focus on half-perimeter wirelength (HPWL…
-
New defense method SRAP improves face-swap protection with SVD refinement
Researchers have developed SRAP, a novel method for defending against face-swapping deepfakes. SRAP refines adversarial perturbations using singular value decomposition (SVD) and an identity-importance mask. This approa…
-
New method smooths optimization on simplex product spaces
This paper introduces a novel method for optimizing functions on product spaces of simplices, which are relevant to tasks like learning probability distributions and functional data registration. The approach involves r…
-
ImageCLEF 2026: Adversarial Deepfake Generation and Detection Methods Explored
A research paper details a team's participation in the ImageCLEF 2026 Deepfake Detection and Generation Task, employing FLUX.1-dev with PuLID for identity-preserving face synthesis and a multi-model PGD adversarial atta…
-
New algorithm solves quasar-convex optimization with constraints
Researchers have developed a new inexact accelerated proximal point algorithm for quasar-convex smooth functions with general convex constraints. This algorithm achieves an optimal first-order query complexity of $\wide…
-
New RL algorithm PPO-PGDLC enhances policy robustness
Researchers have developed a new reinforcement learning algorithm called PPO-PGDLC, designed to improve policy robustness against uncertainties in transition dynamics. This algorithm integrates Proximal Policy Optimizat…
-
New metric measures how AI security classifier explanations degrade under attack
A new research paper introduces the Explainability Stability Index (ESI) to measure how adversarial attacks affect the explanations of cybersecurity classifiers. The study, which extends prior work to Random Forest and …
-
New detector uses mimetic operators to spot adversarial image attacks
Researchers have developed a new, training-free detector for adversarial image perturbations that utilizes high-order Corbino--Castillo mimetic operators. This detector operates in O(HW) time and does not require access…
-
New research explores multimodal and sparse autoencoder methods to combat LLM jailbreaks
Researchers are developing new methods to combat jailbreaking attacks on spoken language models (SLMs). One approach, JAMA, uses a joint multimodal optimization framework to simultaneously attack both audio and text mod…
-
New AEGIS Framework Enhances Adversarial Detection in Vision Sensors
Researchers have developed AEGIS, a novel framework designed to enhance the robustness of adversarial detection in vision sensor networks. This system integrates a SemantiGAN module for semantic discrimination of incons…
-
Quantum neural networks use noise for robust intrusion detection · arXiv research
This paper introduces a rigorous theoretical framework for stochastic quantum neural networks (SQNNs) to enhance adversarial robustness in network intrusion detection. The research proposes a "decoherence-contraction th…
-
New Veriphi System Integrates Attacks and Certification for Neural Network Verification
Researchers have developed Veriphi, a new system for verifying neural networks that integrates fast adversarial attacks with formal bound certification. Experiments on MNIST and CIFAR-10 datasets revealed that the effec…
-
Withdrawn paper reveals substrate-dependent adversarial failure in AI models
A research paper, now withdrawn, explored adversarial robustness in object detectors, specifically focusing on a phenomenon termed "Quality Corruption" (QC). The study observed that one model, EMS-YOLO, a spiking neural…