A novel method for implementing AI attention mechanism using the natural synchronization of physical oscillators, enabling more energy-efficient computation.

Traditional genome editing relying on double-stranded DNA breaks (DSBs) can cause unintended indels, translocation events, and cytotoxic responses in target cells. To overcome these limitations, a precise site-specific modification technology is required that operates without inducing double-stranded DNA cleavage. Developed by UC Berkeley researchers, this technology utilizes engineered fusion proteins comprising a Cas domain (such as a Cas9 nickase or dCas9) integrated with a nucleic acid-editing enzyme domain, such as a deaminase. The molecular mechanism leverages the RNA-guided targeted binding capability of Cas9 to localize the enzymatic editing domain to a specific genomic locus, facilitating direct nucleotide alterations (e.g., cytidine or adenosine conversions) without introducing double-stranded cuts. This system delivers high-precision target modification, significantly reduces off-target insertion/deletion mutations, and improves cell viability during genomic modifications compared to conventional double-strand cleavage mechanisms.

A novel mm-wave transmitter architecture that integrates load-modulated balanced power amplifiers (LMBPA) and double quadrature (DQ) up-conversion for enhanced efficiency, image suppression, and antenna resilience.

A compact, low-power frequency modulated continuous wave (FMCW) radar system enabling high-resolution direction-of-arrival (DoA) detection using a single leaky-wave antenna and transceiver chain

An advanced microfluidic platform that enables high-throughput 3D imaging of live, non-adherent single cells using droplet-based microvortex rotation.

Researchers at the University of California, Davis have developed novel herbicidal and plant growth regulating compositions effective against herbicide-resistant weeds and suitable for use in crop management.

A pioneering framework that enables formal verification of Large Language Model (LLM) compliance with specified interaction protocols using natural language and Finite State Machine (FSM) structures. It is complemented by methods for rapid protocol development and scalable multi-agent coordination.

Brief description not available