A Mixture-of-Experts (MoE) architecture is a type of neural network design that uses multiple specialized models - called “experts” - and dynamically selects which ones to activate for each input. Instead of having all parts of a model process every piece of data, MoE allows only a few experts to...

Internal Coherence Maximization (ICM) is a method for fine-tuning language models without needing any human-provided labels. Developed by Anthropic, it helps a model improve itself using only unlabeled data by leveraging its own ability to find the most internally consistent answers. In...

Google’s AI Edge Gallery is an experimental, open-source Android app that lets you run generative AI models directly on your phone - without needing an internet connection once the models are downloaded. It’s part of a broader shift toward edge AI, where the computation happens on-device instead...