Building Trustworthy AI: Combining Simple Models for Robust Performance
A new approach leverages mixtures of transparent local models to create interpretable machine learning systems with guaranteed performance bounds.
Science
Can AI Truly Hear? Testing Audio Models in the Real World
A new benchmark reveals that current audio AI struggles with the complexities of everyday sound, often performing worse with common noise reduction techniques.
The Benefits of Being Forgetful: How AI Can Learn Like Humans
New research reframes forgetting in large language models as a crucial cognitive process, not a limitation, and demonstrates a method for leveraging this to improve reasoning abilities.
Taming the LSST Data Flood: Meet Alertissimo
The Vera C. Rubin Observatory’s LSST will generate an unprecedented deluge of astronomical alerts, and a new tool called Alertissimo is designed to help scientists manage and analyze this real-time stream.
Predicting Winter Surges: A New Alert System for Pediatric Respiratory Illness
![The study demonstrates a predicted hospitalization curve [latex] H_{\text{SIR}}(t) [/latex], revealing a peak magnitude [latex] h_{\text{SIR}} [/latex] occurring on day [latex] t_{\text{SIR}} [/latex], thereby establishing a quantifiable relationship between epidemiological parameters and peak healthcare demand.](https://arxiv.org/html/2601.09821v1/figs/ModelSIR.png)
Researchers have developed a forecasting model to anticipate peaks in pediatric respiratory infections, offering hospitals crucial time to prepare for increased demand.
Seeing is Understanding: AI Gains Smarter Vision for Roadside Assets
A new approach harnesses the power of artificial intelligence to accurately identify and analyze roadside infrastructure, paving the way for proactive maintenance and improved urban planning.
Decoding Nature’s Rhythms with Machine Learning
New data-driven methods are allowing scientists to unravel the complex dynamics of ecological and epidemiological systems with unprecedented accuracy.
Unlocking Black Box Time Series: A New Approach to Explainable AI
Researchers have developed a novel framework for dissecting complex time series models and revealing the underlying drivers of their predictions.
When AI Plays Doctor: Safety Advances in Child Health
New research reveals that large language models are becoming safer in simulated pediatric consultations, but bigger isn’t always better.
Guarding the Gate: Probing for Safer Language Model Responses
![The reduction in overlap between density distributions of original and Safety-Awareness Enhanced [latex]\mathcal{L}_{disc}[/latex] on both benign and harmful samples-achieved through in-decoding probing-indicates the method effectively isolates signals indicative of harmful content, suggesting a robust mechanism for discerning potentially dangerous inputs.](https://arxiv.org/html/2601.10543v1/x6.png)
Researchers have developed a novel technique to bolster the defenses of large language models against adversarial prompts designed to bypass safety protocols.