🏢 Samsung AI Cambridge

QBB: A novel post-training quantization method for LLMs dramatically improves efficiency by replacing multiplications with summations, achieving state-of-the-art results with minimal accuracy loss.