Researchers developed physics-embedded inverse learning (PEIL), which uses a fixed, differentiable physical solver to turn signal-recovery error into supervision for parameter estimation without parameter labels. In high-mobility wireless tests, PEIL generalized to unseen channel profiles and speeds without retraining and matched a parameter-supervised baseline using one twentieth as many training samples. PEIL was also used to estimate coil sensitivity maps for parallel MRI.

A research team led by faculty at the Duke Quantum Center (DQC) has observed string-breaking dynamics related to particle-antiparticle formation on a quantum simulator, one of the first in the quantum-physics field.

New York University researchers have trained an AI model to learn chemical patterns associated with stability in drug-like molecules and accurately predict where their hydrogen atoms should be positioned.

Their research, published in the journal Chemical Science, addresses a longstanding challenge in molecular design and drug discovery: how to rapidly and reliably determine the stable form of molecules that share a molecular formula but readily convert into different forms. 

The novel physics-constrained AI framework PHICS realizes a full closed loop spanning multiscale inverse design and experimental validation. Leveraging interface, phase, and carrier engineering, PHICS breaks long-standing material trade-offs to yield composite phase-change materials that achieve simultaneous enhancements in thermal conductivity and electrical insulation while maintaining high thermal storage density, paving new pathways for designing advanced electronic thermal management materials.

UCLA researchers developed a scalable, energy-efficient optical-neural system for deepfake video detection. The system combines a lightweight digital front-end with a spatially-multiplexed optical decoder, simultaneously analyzing ≥15 video streams in a single optical pass. On the Celeb-DF dataset, it experimentally achieved accuracy/sensitivity/specificity of 97.79/99.86/95.72%, respectively. Researchers also demonstrated robustness against video degradation, noise, compression, misalignments, and adversarial attacks, highlighting the potential of optical computing for high-throughput and multiplexed content moderation and security-critical AI applications.

Until now, organic light-emitting diodes have not been compatible with the high-precision, but chemically harsh manufacturing processes used in the semiconductor industry. Researchers at ETH Zurich and the University of Helsinki have developed light-emitting polymers that also serve as photoresists and can be precisely patterned using photolithography – an established process in the chip industry. This enables the production of tiny, coloured OLED pixels for screens, as well as precise light sources for research and medical technology.

Researchers have developed a fast, non-intrusive diagnostic approach for resolving the electron-density evolution of plasma generated by microscale explosive detonation. By combining microwave Rayleigh scattering with high-speed imaging, optical emission spectroscopy, thermodynamic calculations, and finite-element simulations, the study quantitatively determines the time-resolved electron density of HMX detonation plasma. The electron density reaches the order of 10²⁵ m⁻³, persists for approximately 1.91 μs, and exhibits a pronounced multi-peak evolution associated with the pulsation of detonation products.