Predicting the Future, with Confidence
A new neural network framework moves beyond single-value predictions to model the full range of possible outcomes in complex dynamical systems.
Science
Predicting the Future of AI Compute: A New Approach to GPU Workload Forecasting
Accurately forecasting demand for GPUs is critical for efficient AI infrastructure, and researchers have developed a new framework that dramatically improves prediction accuracy.
Building Trust in Collaborative AI
A new approach to federated learning empowers industrial networks to proactively manage trust and improve system stability.
Scaling Up Weather Prediction with Neural Networks
A new study reveals how systematically increasing the size of neural networks, combined with strategic training techniques, can dramatically improve weather forecasting accuracy.
Seeing is Believing: AI-Powered Crop Mapping with Transparent Insights
A new framework combines the power of deep learning with explainable AI to achieve near-perfect crop classification and provide farmers with actionable data.
When Less Data Means More Resilience
![Epistemic compression fundamentally differs from likelihood maximization by employing rate reduction-quantified as [latex]\Delta R[/latex]-as a geometric sieve that collapses high-variance noise onto the underlying low-dimensional manifold, effectively orthogonalizing class subspaces and recovering the invariant causal structure, in contrast to the unconstrained feature space and brittle representations produced by fitting noise instances to separate classes.](https://arxiv.org/html/2603.25033v1/figures/fig_whitebox_geometry.png)
As AI models grow in complexity, a surprising strategy for improving performance in unpredictable conditions is gaining traction: deliberately limiting the information they process.
Ripples of Risk: Forecasting Market Shocks with Spillover Effects
![The analysis of spillover components-[latex]q^{s}_{i,t}[/latex]-across several models-SAV CAViaR-SE, AS CAViaR-SE, and IG CAViaR-SE-from March 19, 2008, through November 14, 2025, demonstrates how systemic risk propagates through interconnected financial instruments, revealing distinct patterns of influence for each modeled approach.](https://arxiv.org/html/2603.25217v1/x5.png)
A new modeling approach enhances the accuracy of Value at Risk and Expected Shortfall predictions by accounting for how risk spreads between assets.
Mapping Disease Spread with the Power of Networks
![The model’s sensitivity was assessed by quantifying the loss in balanced accuracy-measured on test host pairs-following the random permutation of dataset variables, revealing that the [latex]Genetic\_Distance[/latex] attribute holds particular importance, though it is exclusively defined for training edges within the Graph Neural Network, while other significant variables function as node-level attributes.](https://arxiv.org/html/2603.24745v1/importance_boxplot_Cumbria.png)
New research demonstrates how graph neural networks can unlock deeper insights from genomic and relational data to improve our understanding of infectious disease outbreaks.
The Art of the Prompt: Measuring AI’s Potential for Manipulation
A new framework evaluates how easily large language models can be used to influence and deceive, revealing critical variations in their manipulative capabilities.
Connecting the Dots in Emergency Response
A new multi-agent system aims to transform fragmented 9-1-1 data into a unified and evolving picture of emergency incidents.