As of now, tank trucks in Mannheim can for the first time load fuel blends that not only contain bioethanol—common in the familiar E10 and E5 grades—but also other renewable components such as electricity-based fuels or biogenic fuels, known as reFuels. A new pilot plant at the tank farm of Exolum Mannheim GmbH flexibly mixes the required proportions and documents the greenhouse gas savings digitally for every delivery. Karlsruhe Institute of Technology (KIT) provides scientific support for the construction of the plant

Hydrogen fuel cell heavy-duty trucks offer a cleaner alternative to diesel transport, but public support is essential for large-scale adoption. In a new study, researchers surveyed households in South Korea to measure willingness to pay for expanding hydrogen truck deployment. The results show strong public acceptance, with benefits exceeding carbon reduction costs, indicating the policy is socially profitable and supports long-term low-carbon transport transitions under national climate policy goals frameworks.

Dramatically reducing energy consumption while accelerating the processing of large amounts of data. This is the aim of the new chip developed by a group of researchers from the Department of Electronics, Information and Bioengineering – DEIB at the Politecnico di Milano, led by Professor Daniele Ielmini, and presented in the study published in the prestigious journal Nature Electronics, with the researcher Piergiulio Mannocci as the first author.

Researchers developed ShapKAN, a deep learning model integrated into the AI4Min-PE platform (http://pe.ai4mineral.com), enabling instant prediction and visualization of key thermodynamic parameters up to 500 GPa. This open AI tool supports the discovery of new chemical behaviors of minerals and elements under extreme conditions.

High-density tactile sensor arrays are essential for next-generation electronic skins, yet increasing array density inevitably amplifies lateral strain–induced mechanical crosstalk, which severely undermines signal stability, recognition reliability, and long-term service performance—particularly in dynamic applications such as Braille reading. Here, researchers report a capacitive tactile array incorporating a gradient-modulus strain isolation layer that decouples mechanical functions along the thickness direction, effectively suppressing nonlocal strain transfer while maintaining high sensitivity. Guided by quantitative structural design principles supported by simulation and experiment, the device achieves a sensitivity of 3.92 kPa⁻¹ with a crosstalk coefficient of only 4.39%. Integrated into a “sliding–recognition–speech” Braille platform, it achieves 100% static and 99% dynamic recognition accuracy while reducing false triggers by over 70%. This work demonstrates that structural suppression of mechanical crosstalk is a powerful pathway to ensuring the stable, reliable operation of high-density electronic skins and offers a scalable design strategy for future assistive tactile perception technologies.

Recently, a research team led by Professor Zhang Jiachao from the College of Food Science and Engineering, Hainan University, published a new study in Science Bulletin entitled “Calenduloside E produced by Bifidobacterium animalis B960 increases glibenclamide efficacy and alleviates drug-induced adverse effects in a type 2 diabetes model.” This study directly addresses the long-standing challenge of the “microbial drug black hole” in the pharmacotherapy of metabolic diseases. The team successfully identified a tropical probiotic strain with dual “efficacy-enhancing and toxicity-reducing” properties, Bifidobacterium animalis subsp. lactis B960 (B960). For the first time, the researchers systematically elucidated the molecular mechanism by which this strain produces the functional metabolite Calenduloside E to synergistically enhance the therapeutic efficacy of glibenclamide while mitigating its adverse effects. Importantly, these findings were further validated using a human-derived gut–liver interconnected organ-on-a-chip platform that recapitulates key aspects of the human physiological microenvironment in vitro. Overall, this work provides an innovative strategy for the development of probiotic-based adjuvant therapies for the treatment of type 2 diabetes.

Large language models (LLMs) can be manipulated by "trick prompts" that try to override safety rules. We propose a lightweight defense that improves detection while avoiding needless refusals of harmless requests. Two LLM agent teams-one generating tougher test prompts and one analyzing them-iterate in a feedback loop to sharpen judgments. Instead of costly fine-tuning, the method learns through in-context logs, steadily boosting accuracy across standard metrics.