Hijacking AI Assistants: A New Era of Attack
Researchers have demonstrated a powerful new method for manipulating AI agents, exposing vulnerabilities beyond typical prompt injection defenses.
Science
Fragile AI: Can Mamba Handle the Pressure in Medical Imaging?
![Vision Mamba proposes a selective state space model-a departure from traditional transformers-that utilizes a hardware-aware, temporally selective scan to process visual inputs, achieving a linear scaling in sequence length and promising improved efficiency in vision tasks through a principled reduction of computational redundancy inherent in attention mechanisms [latex]O(N) [/latex] versus [latex]O(N^2)[/latex].](https://arxiv.org/html/2602.16723v1/Fig/visionMamba.jpg)
A new study reveals that Mamba-based medical image analysis, despite its impressive performance, is surprisingly susceptible to both malicious attacks and real-world hardware failures.
Beyond Einstein: Mapping Multigravity with Black Hole Solutions
New research leverages the Kerr-Schild approach and the double copy to construct exact solutions in multigravity theories, revealing unexpected links to fundamental field interactions.
Beyond Accuracy: Measuring the true Strength of AI Reasoning
New research reveals that achieving high scores on reasoning tasks doesn’t guarantee a system’s ability to apply or justify its logic.
Black Hole Ripples: Unlocking the Secrets of Gravity’s Edge
New research reveals how the fundamental properties of black holes dictate their response to external forces, linking horizon behavior to observable tidal distortions.
Unlocking Meson Mass Patterns: A New Look at Chiral Symmetry
![Dynamical chiral symmetry breaking, as proxied by the quark mass function [latex]M(0)[/latex], exhibits a transition-sharply defined in the chiral limit-to a crossover behavior with non-vanishing quark masses, a phenomenon quantified by the parameter [latex]\mathcal{M}[/latex] derived from parabolic fits avoiding singularities at a given coupling strength [latex]D_{I}[/latex].](https://arxiv.org/html/2602.17456v1/x4.png)
Research reveals a connection between the internal structure of quarks and the observed degeneracy of meson masses, challenging conventional understandings of chiral symmetry restoration.
Can AI Agents Be Hacked Over Time?
New research reveals that even sophisticated AI agents are vulnerable to subtle, long-term attacks that exploit extended interactions.
Boosting Wireless Security with Intelligent Surfaces
This review explores how reconfigurable intelligent surfaces can be strategically deployed to maximize data security in modern wireless communication networks.
Verifying AI: A Faster Path to Trustworthy Machine Learning
New research introduces Jolt Atlas, a system designed to dramatically accelerate the process of verifying machine learning models using cryptographic techniques.
Hidden Signals: Injecting Persistent Backdoors into AI Vision-Language Models
Researchers have demonstrated a new method for embedding subtle, yet enduring, backdoors into artificial intelligence systems that process both images and text.