



Mila - Institut québécois d'IA
4.1K posts

@Mila_Quebec
Le plus grand centre de recherche universitaire en apprentissage profond. The largest academic research center in deep learning. 🦋@mila-quebec.bsky.social









New paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
















Your LLM already knows the answer. Why is your embedding model still encoding the question? 🚨Introducing LLM2Vec-Gen: your frozen LLM generates the answer's embedding in a single forward pass — without ever generating the answer. Not only that, the frozen LLM can decode the embedding back into text. 🏆 SOTA self-supervised embeddings 🛡️ Free transfer of instruction-following, safety, and reasoning















