Testing AI’s Boundaries: Risks and Realities of Agentic Systems
A large-scale international evaluation reveals significant vulnerabilities in AI agents tasked with complex actions, highlighting critical gaps in safety methodologies.
Science
Crypto’s Volatility Problem: Why Standard Risk Models Fail
New research reveals that commonly used financial models underestimate cryptocurrency risk, potentially leaving investors unprepared for significant losses.
Decoding Corporate Risk with AI
![Sectoral distinctions in company risk profiles emerge organically from the data, as demonstrated by the substantial divergence in similarity distributions-companies sharing two-digit Standard Industrial Classification codes exhibit markedly higher risk profile similarity [latex] (5,263 pairs) [/latex] than those in different industries [latex] (101,228 pairs) [/latex], despite the taxonomy mapping process being explicitly devoid of industry-specific information.](https://arxiv.org/html/2601.15247v1/x1.png)
A new approach uses artificial intelligence to automatically identify and categorize potential risks disclosed in company filings.
Taming the Alert Storm: Smarter Monitoring for Cloud Systems
As cloud infrastructure grows, so does the deluge of alerts, demanding intelligent systems to prioritize critical issues and reduce operator burnout.
Divide and Conquer: Building More Reliable AI with Organizational Principles
New research shows that structuring AI systems like human organizations – with specialized roles and independent checks – dramatically improves their performance and error detection.
Swarm Intelligence Takes Flight: AI-Powered UAVs at the Edge
![UAV swarm architectures are explored through three distinct deployments-standalone, edge-enabled, and edge/cloud-enabled-each offering varying levels of computational resource allocation to individual agents and collectively enabling a range of applications dependent on optimized agent behavior as defined by [latex] agent_{i} [/latex].](https://arxiv.org/html/2601.14437v1/x1.png)
A new approach combining agentic artificial intelligence and edge computing is enabling more scalable and resilient autonomous operation for drone swarms.
Buffering the AI Boom: A New Approach to Power Grid Stability
As demand from artificial intelligence data centers surges, a novel optimization framework aims to bolster grid resilience against unpredictable load fluctuations.
Building AI You Can Trust: Safeguarding Large Language Models
As large language models become increasingly powerful, ensuring their responsible development and deployment is paramount.
Beyond Resolution: Neural Networks That Understand Scale
New research demonstrates how incorporating scale invariance into neural network design enables robust extrapolation to unseen data scales, unlocking better modeling of self-similar phenomena.
Forecasting Cyberattack Surges: A New Approach to IDS Alert Analysis
Researchers have developed a method to predict escalating periods of high-intensity network intrusion attempts by analyzing trends in security alert streams.