Subtle Sabotage: Crafting Hidden Attacks on Graph Neural Networks
Researchers have developed a new technique to subtly manipulate graph neural networks, creating backdoor vulnerabilities that are difficult to detect.
Science
Buildings That Listen: Mapping Footsteps with Structural Vibration
![A building’s structure functions as a reservoir computer, localizing footsteps by converting the mechanical impulses of walking into dispersive vibrational fields sampled by implanted accelerometers, then projecting these signals-normalized and reduced via Principal Component Analysis-into reservoir state vectors used with trained weights to accurately estimate footstep location [latex] \hat{\mathbf{z}}\_{k}=(\hat{x}\_{k},\hat{y}\_{k}) [/latex].](https://arxiv.org/html/2603.04610v1/2603.04610v1/x1.png)
New research demonstrates that a building’s own structure can be harnessed as a sensor network to pinpoint footstep locations within its walls.
AI on the Stand: Can Algorithms Uncover Sentencing Bias?
A new study investigates whether artificial intelligence can help legal professionals build robust statistical evidence to support claims of racial disparities in sentencing.
Smarter Turbulence Models: Harnessing Data and Symmetry
![At an initial time, Large Eddy Simulation (LES) solutions align with filtered Direct Numerical Simulation (DNS) data; however, for the Clark model, this alignment breaks down around [latex]t = 2.1[/latex], leading to computational instability and preventing solution completion.](https://arxiv.org/html/2603.05325v1/2603.05325v1/figures/snellius/velocities-z.png)
New research explores how data-driven approaches, particularly those respecting fundamental physical symmetries, can improve the accuracy of large-eddy simulations.
Shrinking the Spectrums: Efficient Deep Learning for Hyperspectral Imagery
![The research details the implementation of convolutional neural networks across varying dimensionalities-one-dimensional for spectral analysis [latex]CNN1D[/latex], two-dimensional for spatial analysis [latex]CNN2D[/latex], and three-dimensional to integrate both spectral and spatial information [latex]CNN3D[/latex]-demonstrating a progression in model complexity for comprehensive data interpretation.](https://arxiv.org/html/2603.04720v1/2603.04720v1/cnn.png)
A new study systematically evaluates techniques to compress deep learning models, making them more practical for land cover classification using complex hyperspectral data.
Predicting Nuclear Reactions with Confidence
![The BNN-I6 model accurately predicts (n,p) reaction cross sections across a wide range of nuclei, with root-mean-square deviations [latex]\sigma_{rms}[/latex] generally remaining small-indicating successful capture of the systematic dependence on neutron and proton number-though slightly larger deviations emerge for mid-mass and heavy nuclei, potentially due to increased structural complexity and limited experimental data in those regions.](https://arxiv.org/html/2603.04789v1/2603.04789v1/x6.png)
A new machine learning approach leverages Bayesian Neural Networks to accurately forecast neutron-induced reaction probabilities and, crucially, quantify the uncertainty in those predictions.
Seeing the Future: AI Predicts Rainfall with Greater Accuracy
A new deep learning model leverages multimodal weather data and physical principles to significantly improve short-term precipitation forecasting.
Radar-Guided Jamming: Securing Wireless Communications with Sensing
A new framework uses radar sensing to intelligently disrupt eavesdropping attempts, enhancing the security of integrated sensing and communication systems.
Decoding Deception: An AI That Helps You Spot Manipulation
Researchers have developed an agentic AI framework that moves beyond simple detection to offer users self-discovery prompts when manipulative communication patterns are identified in ongoing conversations.
Pinpointing Policy Impacts: A New Bayesian Approach
A novel statistical method offers a more robust way to evaluate the effectiveness of climate policies amidst complex and shifting economic landscapes.