Shrinking AI memory boosts accuracy, study finds
Reports and Proceedings
Updates every hour. Last Updated: 10-Jun-2026 08:16 ET (10-Jun-2026 12:16 GMT/UTC)
Researchers have developed a new way to compress the memory used by AI models to increase their accuracy in complex tasks or help save significant amounts of energy.
Experts from University of Edinburgh and NVIDIA found that large language models (LLMs) using memory eight times smaller than an uncompressed LLM scored better on maths, science and coding tests while spending the same amount of time reasoning.
This study shows that CRISPR-induced mutation of the mstnb gene in grass carp significantly accelerates growth. Edited fish were longer, heavier, and had thicker bodies than controls, mainly due to increased muscle fiber number rather than fiber size. The results demonstrate that targeting myostatin gene (mstnb) can effectively enhance muscle development and growth in aquaculture species.
An engineering team at Jiangxi Normal University, in collaboration with South China University of Technology, reports an entropy-driven strategy to construct low-topology-entropy silicone elastomers (LTE-SEs) in the journal Wearable Electronics. The materials achieve ultra-softness and ultra-high stretchability while maintaining high strength, and are successfully applied in skin-conformal flexible encapsulation, UV-protection patches, and fully encapsulated safety-positioning insoles, providing a new generation of substrate materials for long-term, comfortable, and reliable wearable electronics.