LATMiX: Learnable Affine Transformations for Microscaling Quantization of LLMs

LATMiX introduces learnable affine transformations to minimize quantization errors in LLMs, offering a theoretically rigorous approach to microscaling that s...

Level: advanced

By Ofir Gordon, Lior Dikstein, Arnon Netzer, Idan Achituve, Hai Victor Habi

Category: research