Training-free framework accelerates large language models without retraining
Peer-Reviewed Publication
Updates every hour. Last Updated: 30-Oct-2026 10:16 ET (30-Oct-2026 14:16 GMT/UTC)
Researchers from the University of Osaka, Gunma University and Kyushu University created lipid mimic molecules that form “island”-like groups called lipid domains in artificial lipid membranes. The domains are thermally stable because the propellor-shaped moiety of the lipid mimic molecules interlock and the surrounding lipids fit into the spaces between. This new method of forming domains may allow artificial lipid membrane materials to be created for catalysis, capturing molecules, and drug delivery.
Scientists have engineered tobacco and lettuce plants that carry the gene for the protein myoglobin, a molecule abundant in animal muscle fibers, in the genome of their chloroplasts. This makes it possible to mass-produce myoglobin in crops to give a meaty flavor and color to meat alternatives. The expected yield at scale could rival those from livestock farming, while greatly reducing water use and greenhouse gas emissions.
Glioblastoma is difficult to treat because tumour cells infiltrate the surrounding brain tissue and the blood-brain barrier limits how well drugs and radiotherapy reach the brain. Researchers from the University of Technology Sydney (UTS), Harvard and Henan universities have discovered a ‘double-punch’ nanozyme platform designed to address both problems with a single tool of smart nanoparticles.
- A research team led by Professor Mooseok Jang of the Department of Bio and Brain Engineering has developed a technology that reconstructs a transparent object hidden between two dynamic scattering layers—layers that scatter light in multiple directions—from a single shot.
- The technology is expected to have applications in a wide range of fields, including precision inspection for semiconductors and displays and biomedical imaging.
AI systems increasingly influence how people search for information, receive recommendations, learn, work, and create. But these systems are only as reliable as the quality of the data used to train them. Supported by an NSF CAREER grant, Jiaqi Ma hopes to develop tools to make data curation more principled and transparent and improve the performance, reliability, and safety of widely used technologies such as language models and recommendation systems.