<ul><li>Large Language Models (LLMs) face challenges in edge deployment due to their massive parameter scale.</li><li>Vector Quantization (VQ) is a prevalent solution for quantizing LLMs at low-bit with considerable accuracy.</li><li>Polar Coordinate Decoupled Vector Quantization (PCDVQ) proposes independent quantization of direction and magnitude parameters for better accuracy.</li><li>Experimental results show PCDVQ outperforms baseline methods at 2-bit level by at least 1.5% zero-shot accuracy.</li></ul>

PCDVQ: Enhancing Vector Quantization for Large Language Models via Polar Coordinate Decoupling

Discover more