TurboQuant (ICLR 2026) vector quantization for memory/RAG embedding compression | 5-8x压缩 98%+召回率 | numpy only, no GPU

A zero-dependency, BP-free Forward-Only Neural Network using Dual-Rail Positive ($\mathbb{R}^+$) logic and Hadamard Gating. Where the world is the training set and updates are the Logos.2. 標籤 (Topics)