A U.S. federal agency is investing $38 million in a provocative and potentially transformative project: AI-powered medical "digital twins" to help doctors personalize treatments for critical ill patients in hospitals. The project, led by a U. of Vermont surgeon and researcher, will help doctors treat the most challenging illnesses in the ICU—sepsis, burns, severe trauma, and disordered immune responses. The funding is from ARPA-H in the U.S. Dept. of Health and Human Services and is the largest research award in UVM's history.

A new piece of equipment has been added to the optics toolkit. Physicists at the University of Oldenburg, Germany, have succeeded in generating 3D light fields. Using this method, they have placed electrons into quantum states that were previously inaccessible through experimentation. In the journal Physical Review Research, the team writes that the method is particularly well-suited to studying molecules that exist in two mirror-image variants, so-called chiral molecules.

Researchers at Fudan University have developed a crosslinked potassium polyacrylate skin membrane that lets microrobots generate electricity by walking on metal or silicon surfaces. The membrane uses atmospheric moisture and oxygen to convert environmental chemical energy into power, achieving 133 mW cm⁻² on zinc. It also senses substrate material and roughness, enabling active energy pursuit. The system powered over 10,000 walking steps and remained stable after 1,000,000 steps.

Researchers found that carefully balancing metallic cobalt and cobalt oxide enables highly selective electrocatalytic hydrogenation of nitrogen-containing compounds. The catalyst converted pyridine to piperidine with a more than 99% yield, while intermittent electrolysis maintained its optimal chemical state, offering a potential route toward sustainable chemical manufacturing without use of scarce precious metals.

A “hidden variable” that can determine the success or failure of a chemical reaction has been found in the immediate vicinity of electrodes.

The valence-electron concentration has been conventionally used to identify and classify magnetic ground states in various intermetallic compounds. In quasicrystal approximants, the predictive power of valence-electron concentration is found to be limited across different alloy families and constituent elements. Now, researchers establish the lattice parameter as an experimentally accessible, unified structural descriptor for magnetic ground state selection in Tsai-type icosahedral compounds. This breakthrough will help identify novel quasicrystals and approximant crystals with emergent magnetic properties. 

The research team from National University of Defense Technology systematically described the memory efficiency issues in the training process of large language models (LLMs) with Transformer architecture for scientific fields. As the scale of models continues to expand, memory requirements have become a key bottleneck restricting their applications in scientific fields such as biology, medicine, chemistry, and meteorology. The article systematically reviewed memory-efficient pre-training technologies at the algorithm level, system level, and software-hardware collaborative optimization, and combined with the AlphaFold 2 case, demonstrated how customized memory optimization methods can reduce storage overhead while maintaining prediction accuracy.

A new study published in Advances in Atmospheric Sciences shows that greening in China generally cools the local climate with more rainfall over southern and northeastern China, while greening in India tends to produce warming and less rainfall. The Co-occurring greening in China and India yields uniform climate feedbacks across over half global lands, producing cooler and wetter southern China, cooler western Eurasia, and warmer northern North America.