MIT's MeMo lets teams swap in a better LLM without retraining — and performance jumps 26%

MIT's MeMo keeps AI memory separate from reasoning, so teams can upgrade their LLM without retraining and see a 26% performance gain, researchers say.

Axial encoding unlocks up to eightfold faster 3D microscopy with less light

A research team from HKU Engineering has pioneered a fundamentally new imaging strategy known as AIMED (Arbitrary illumination microscopy with encoded depth), which utilizes a sub-sampling approach.