Accurately predicting passenger flows is essential for preventing congestion and delays in subway networks. In a recent study, researchers developed a multi-time-scale model that uses recurring day-to-day and week-to-week travel patterns to predict both passenger destinations and travel times one day ahead. Tested on Shenzhen subway data, the model outperformed several widely used forecasting methods across most stations.

Aqueous Zn-organic batteries are promising due to the sustainability and tunability of organic cathodes. However, the critical challenge in their practical application lies in dissolution, degradation, and sluggish kinetics, ultimately degrading the cycling stability and rate capability. Herein, we demonstrate a simple effective electrolyte engineering strategy by introducing potassium chloride (KCl) as a co-solute into the ZnCl₂ electrolyte to achieve high-performance Zn batteries utilizing 3,4,9,10-perylenetetracarboxylic dianhydride (PTCDA) as the cathode. Experimental studies and molecular dynamics simulations reveal that KCl addition not only enhances electrolyte conductivity but also modulates Zn²⁺ solvation environment to form [Zn(H₂O)₂Cl₄]²⁻, significantly improving ion diffusion kinetics. Consequently, the Zn//PTCDA in the ZnCl₂-KCl electrolyte exhibits a remarkable capacity of 124.7 mAh g⁻¹ with an average voltage of 0.65 V, exceptional rate performance (56% capacity retention at 30 A g⁻¹), and prolonged cycling performance (90.9% retention after 10,000 cycles). Experimental and density functional theory mechanistic studies unveil a new reversible Zn²⁺/K⁺ co-storage mechanism in the PTCDA cathode, where K⁺ acts as a charge shield and structural pillar, synergistically decreasing the ion migration energy barrier, enhancing reaction kinetics, and stabilizing the cathode structure. This work elucidates dual-ion storage chemistry and highlights rational electrolyte design for durable, high-power metal–organic batteries.

A research team led by Professor Jiandong Ding at Fudan University has derived a set of equations for random degradation inspired by the classic Poisson raindrop problem, providing the first theoretical framework to quantify “acquired heterogeneity” — the uneven degradation that emerges during service rather than from manufacturing defects. Applying this framework to biodegradable metal stents, the team discovered that coating iron with polylactide (PLA) increases the corrosion rate by 3-fold but reduces corrosion inhomogeneity by 4,000-fold in blood-mimetic medium, paradoxically leading to later device fracture despite faster degradation. The resulting metal–polymer composite stent was validated in preclinical porcine studies, clinical trials involving 1,108 patients across multiple cohorts, and 5-year follow-up of 45 first-in-human cases, demonstrating safety, durability, and reliability. The theoretical framework is, in principle, applicable to any field where random spatiotemporal variability undermines system performance, from ship corrosion and bridge fracture to climate variability and big data analytics.

The FAST All Sky HI Survey (FASHI) team has released its second data set (DR2), cataloging 156,411 extragalactic neutral hydrogen (HI) sources over ~19,500 square degrees — nearly half the entire sky. The number of detected sources is nearly four times that of the first data release (DR1) and five times that of the Arecibo ALFALFA survey. With a median sensitivity of 0.57 mJy/beam, FASHI DR2 represents the largest and most sensitive blind HI survey ever conducted, providing a landmark dataset for studies of galaxy evolution, cosmology, and large scale structure in the local Universe.

A research team at Southern Medical University reveals a complete pathway running from gut microbiota to liver detoxification enzymes: depression disrupts the *Clostridium*-LCA-PXR-CYP3A11 metabolic axis, causing the gelsemium toxin to accumulate in the body and sharply worsening neurotoxicity

An efficient few-shot framework was developed for large-scale annotation of light-trapped insect images. MLTIDD integrated Grounding DINO, SAM, and SAHI for multi-scale insect detection, while InsectSSRL exploited massive unlabeled insect images to enhance few-shot classification. MLTIDD achieved 79.6% AP50–95 and 90.8% AR, while InsectSSRL achieved 80.35% classification accuracy using only 5% labeled data. The framework reduced expert annotation time by approximately 80%, enabling efficient large-scale dataset construction.

Researchers at Wuhan University developed a molecular-sieve-assisted confined molten-salt route to grow 2.66 nm Na₂Nb₄O₁₁ single-crystal sheets. Devices made from the material combined sensitive ultraviolet detection with millisecond response and stable cycling, and enabled UV imaging and proof-of-concept two-wavelength physical-layer information encoding.

Researchers from Shenzhen University and collaborating institutions have provided a comprehensive review of viologen-linked covalent organic frameworks (Vio-COFs) as an emerging class of charged and redox-active porous materials. With their unique combination of permanent positive charges, reversible viologen redox chemistry, and well-defined porous frameworks, Vio-COFs offer versatile platforms for modulating electron transfer, ion migration, and host–guest interactions. This review highlights recent developments in the design strategies, structural properties, and functional applications of Vio-COFs, emphasizing their potential in catalysis, energy conversion and storage, environmental remediation, and optoelectronics. Overall, the insights presented in this review pave the way for the rational design and engineering of next-generation multifunctional Vio-COF materials.