How AI Might Save Us: Mapping Paths to Survival
A new analysis categorizes potential future scenarios where humanity successfully navigates the risks of increasingly powerful artificial intelligence.
Science
Beyond the Lie: Taming falsehoods in AI Language Models
As large language models become increasingly integrated into critical applications, ensuring factual accuracy and mitigating the risk of fabricated information is paramount.
Beyond the Lab: AI Predicts Concrete Strength with Unprecedented Accuracy
A new study demonstrates the potential of artificial intelligence to rapidly and reliably assess concrete compressive strength, offering a path toward automated quality control in large-scale construction projects.
Tracking the Ripple Effect: Ensuring Reliability in Deep Learning Code
A new framework automatically assesses how changes to deep learning libraries impact the applications that depend on them, preventing silent errors and improving software stability.
Forecasting’s Shifting Sands: A New Approach to Supply Chain Stability
Concept drift – unexpected changes in data patterns – can devastate supply chain forecasts, but a novel framework offers a robust solution for detecting and mitigating these disruptions.
Decoding AI Confidence: A New Approach to Probability Estimation
![PRISM reconstructs a large language model’s probability prediction by aggregating Shapley values-quantifying each factor’s contribution-where factors positively correlated with the outcome are represented in green and those negatively correlated in red, with the size of each representation reflecting the absolute value of its contribution, effectively decomposing [latex] f(\cdot) [/latex] via the sigmoid function [latex] \sigma(\cdot) [/latex].](https://arxiv.org/html/2601.09151v1/x1.png)
Researchers have developed a novel framework that boosts the accuracy and interpretability of how large language models assess the likelihood of their predictions.
Seeing Threats Before They Strike
New research demonstrates how advanced AI can sift through online intelligence to identify potential cyberattacks before they fully materialize.
Echoes of a Colossal Merger: GW231123’s Robust Signal
![Analysis of simulations mirroring the gravitational wave event GW231123, conducted with the NRSur waveform model and assessed through independent detector inference, reveals that divergences between posterior distributions of parameters-including total mass [latex]M_M[/latex], mass ratio [latex]q_q[/latex], luminosity distance [latex]D_{L D}\[/latex], effective inspiral spin [latex]\chi_{eff}[/latex], and precession spin [latex]\chi_p[/latex]-are consistent with expectations for Gaussian noise, with a small percentage of LIGO Livingston and LIGO Hanford pairs exhibiting larger divergences than those observed in GW231123 itself.](https://arxiv.org/html/2601.09678v1/figures_BF/JS_H_L_chip.png)
New analysis confirms the surprisingly high masses and spins observed in the gravitational wave event GW231123 are not artifacts of noise or waveform modeling.
Eyes on the Skies: A New AI for Safer, Smarter Aviation
A novel multimodal foundation model is poised to unify diverse data sources, promising enhanced situational awareness and proactive safety management in civil aviation.
Smarter Flood Forecasting: A New Machine Learning Approach
Researchers have developed a streamlined machine learning framework that significantly improves the accuracy of short-term river flow and peak flood predictions.