A new review maps the complex web of technologies enabling the creation and dissemination of AI-generated sexual abuse material, revealing how efforts to combat it often feel like a futile game of whack-a-mole.
A new framework combines the power of graph-based reasoning with large language models to dramatically improve the analysis and understanding of complex time series data.
A new approach to artificial intelligence mimics the way animals learn to avoid harm, creating more robust and adaptable systems.
![In large language models, dynamic instability during inference-manifested as abrupt shifts in next-token probability distributions and increased uncertainty-can be quantified via a diagnostic signal [latex]I\_t = D\_t + \lambda H\_t[/latex], where [latex]D\_t[/latex] measures distributional shift and [latex]H\_t[/latex] represents uncertainty, with the overall instability strength [latex]S = \max\_t I\_t[/latex] correlating with increased failure risk and the timing of instability episodes offering insights into potential recoverability.](https://arxiv.org/html/2602.02863v1/figures/Fig1.png)
New research reveals a way to detect when large language models begin to falter during reasoning, offering insights into why they sometimes fail.
New research reveals how data analysis can pinpoint the key individuals and automated accounts driving the proliferation of misinformation on X.
A new probabilistic forecasting framework improves power grid resilience by accurately predicting extreme energy demand, even with limited data.
New research details a framework for building and pre-training graph foundation models at an unprecedented scale, unlocking advanced capabilities for heterogeneous graph data.
A new approach combining federated learning and blockchain technology offers a path to more accurate retail demand forecasting and reduced food waste.
Researchers are leveraging the power of large language models to automatically generate complex neural network architectures, creating a valuable resource for improving AI reliability and adaptability.
![Across oceanic basins from 2012 to 2017, a comparative analysis using Convolutional Neural Networks (CNN) and U-Nets demonstrates a strong correlation between remotely sensed chlorophyll saturation [latex]log(ChlSat)[/latex] and reconstructed chlorophyll concentrations [latex]log(Chl)[/latex], indicating the potential for accurate chlorophyll estimation via these distinct deep learning architectures.](https://arxiv.org/html/2602.04689v1/Fig2.png)
New research shows deep neural networks can accurately predict phytoplankton biomass-a key indicator of ocean health-by learning from physical ocean conditions.