OSC: Hardware Efficient W4A4 Quantization via Outlier Separation in Channel Dimension

TL;DR

This paper introduces OSC, a hardware-efficient 4-bit quantization method for large language models that suppresses activation outliers through structured channel separation, improving accuracy and speed.

Contribution

OSC presents a novel outlier suppression framework that combines dual-path computation and structured channel coalescence for efficient low-bit model deployment.

Findings

Achieves only 2.19 and 1.12 point accuracy drops on Qwen models.

Realizes up to 1.78x speedup over W8A8 baseline.

Effectively suppresses activation outliers in 4-bit quantization.

Abstract

While 4-bit quantization is essential for high-throughput deployment of Large Language Models, activation outliers often lead to significant accuracy degradation due to the restricted dynamic range of low-bit formats. In this paper, we systematically investigate the spatial distribution of outliers and demonstrate a token-persistent structural clustering effect, where high-magnitude outliers consistently occupy fixed channels across tokens. Building on this insight, we propose OSC, a hardware-efficient framework for outlier suppression. During inference, OSC executes a dual-path computation consisting of a low-precision 4-bit General Matrix Multiplication (GEMM) path and a high-precision 16-bit branch GEMM path. Specifically, OSC uses an offline group-wise strategy to identify the channels where outliers are located and then performs structured sub-tensor extraction to coalesce these scattered activation channels into a compact dense tensor online. This mechanism implements outlier protection through regularized and high-throughput GEMM operations, achieving a seamless fit with modern 4-bit micro-scaling hardware. Furthermore, for the inputs of W2 where outlier clustering is less pronounced, we integrate a fallback strategy to FP8. Evaluation on Qwen3-8B and Qwen3-30B restricts the average accuracy drop to 2.19 and 1.12 points, respectively. Notably, OSC is highly hardware-friendly, achieving a peak speedup of 1.78x over the W8A8 GEMM baseline on a modern AI accelerator.