Transformers & Reasoning

DeepSeek V3.2 and R1 internals, Multi-head Latent Attention, sparse Mixture-of-Experts routing, and reinforcement learning for reasoning.