Researchers developed DGMoE, a mixture-of-experts framework that better handles individual differences in EEG signals and improves emotion recognition across unseen users. Tested on three public datasets, it achieved state-of-the-art accuracies of 79.5%, 59.1%, and 57.9%, showing a more practical path toward robust affective brain-computer interfaces.

Quantum particles in imperfect or quasiperiodic environments may spread, become localized, or enter critical states with fractal wave functions. These behaviors can combine into seven fundamental localization phases, but a unified framework for all of them has been missing. A team led by Prof. Xiong-Jun Liu at Peking University has now developed a generic spin-1/2 quasiperiodic system that captures all the fundamental phases, proves three theorems, constructs new exactly solvable models, and proposes the implementation with ultracold atoms in optical Raman lattices.

By adapting an algorithm from the 1980s to the modern context of mathematical objects called tensor networks, researchers at the Flatiron Institute extended the reach of classical computation and tackled a class of problems previously claimed to be solvable only by quantum computers.

A recent work published in National Science Review report the first realization of a higher-order Weyl exceptional ring semimetal in a 3D lossy acoustic metamaterial. This study systematically analyzes the non-Hermitian topological properties of such metamaterial, and focuses on two types of higher-order topological phenomena: topological hinge states and hinge-dependent skin effect.