Mapping Crypto Risk: A New Geometric Approach
Researchers are applying techniques from topology to understand the stability of cryptocurrency price movements and improve risk management strategies.
Science
The Algorithmic Pulse: How AI is Shaping Population Health
![Generative AI usage is not uniform across racial and ethnic groups; while Black adults demonstrate the highest rates of application in health-related pursuits (30%) and entertainment (31%), individuals identifying with “Other/2+” ethnicities lead in utilizing the technology for internet searches (54%) and educational purposes (32%), suggesting that assessments of AI exposure require nuanced measurement of <i>how</i> the technology is employed, rather than simply <i>whether</i> it is used-though data from the AmeriSpeak survey ([latex]N=1{,}163[/latex]) indicates some estimates are based on small sample sizes and should be interpreted with caution.](https://arxiv.org/html/2604.14086v1/x2.png)
A new framework proposes applying epidemiological principles to understand and measure the pervasive impact of artificial intelligence on the wellbeing of communities.
Tracing Errors: A New Path to System Reliability
A comprehensive review reveals that combining log analysis with execution trace modeling offers the most effective approach to pinpointing and classifying faults in complex distributed systems.
The Fragile Logic of AI: When Tiny Errors Cause Big Problems
![Despite a vast spectrum of singular values-ranging from [latex]\sigma_1 = 615.3[/latex] to approximately zero-the maximum stable perturbation magnitude remains remarkably consistent at approximately [latex]10^{-{10}}[/latex], demonstrating that instability is a pervasive characteristic of the entire embedding manifold.](https://arxiv.org/html/2604.13206v1/x6.png)
New research reveals that large language models are surprisingly susceptible to numerical instability, potentially leading to unpredictable outputs and systemic failures.
Fighting Fire with Algorithms: A New Approach to Stopping Fake News
Researchers are combining the power of artificial intelligence and social network modeling to develop strategies for curbing the spread of misinformation online.
Predicting IT Chaos: How Machine Learning Tames Change-Related Incidents
A new study demonstrates how machine learning can proactively identify and mitigate IT incident risk stemming from system changes, moving beyond traditional rule-based approaches.
Listening to the Heart: AI Improves Cardiovascular Disease Detection
![A classification model-comprising one- and two-dimensional convolutional layers alongside a Long Short-Term Memory network and trained with the ADAM optimizer-achieved differentiation of five heart valvular conditions using a Gabor dictionary ([latex]\beta=2^{1}[/latex]) and elastic net regularization ([latex]\alpha=0.1[/latex]) across 100,100 experiments, as evidenced by its confusion matrix.](https://arxiv.org/html/2604.12483v1/x14.png)
A new deep learning approach leverages advanced signal processing to enhance the accuracy of heart sound analysis for earlier disease identification.
Can AI Read the Market?
New research rigorously tests whether vision-language models can truly interpret candlestick charts to predict stock price movements.
Uncovering the Laws of Motion: From Chaos to Climate
![Governing equation identification across chaotic systems-specifically the Sprott and Halvorsen systems-demonstrates that performance, measured as the fraction of successful term recoveries from 100 trials, improves with increasing observational data [latex]n[/latex] at a signal-to-noise ratio of 49 dB, and surpasses an 80% success rate threshold for Bayesian-ARGOS, ARGOS, and SINDy methods as the signal-to-noise ratio increases at [latex]n=5000[/latex].](https://arxiv.org/html/2604.11929v1/x11.png)
A new framework efficiently extracts governing equations from complex data, bridging the gap between observation and fundamental understanding of dynamic systems.
When AI Misses, Lives Are on the Line
![The research demonstrates a shift in model behavior-achieved through a proposed method (RCL)-that prioritizes minimizing critical errors in diagnostic accuracy, explicitly moving beyond the high-risk, high-accuracy zone exemplified by Focal Loss and toward a safer operating region where reduced fatal errors outweigh marginal gains in overall [latex]F1[/latex]-Macro score.](https://arxiv.org/html/2604.12693v1/fig2_tradeoff_plot.jpg)
A new approach to machine learning focuses on drastically reducing the most dangerous errors in medical image analysis, prioritizing patient safety above all else.