ARCQuant: Boosting NVFP4 Quantization with Augmented Residual Channels for LLMs

TL;DR

ARCQuant enhances NVFP4 quantization for LLMs by using augmented residual channels, enabling efficient, hardware-friendly inference with minimal accuracy loss and significant speed improvements.

Contribution

It introduces a novel residual channel augmentation technique that maintains a unified NVFP4 format, improving quantization accuracy and hardware efficiency for LLM inference.

Findings

Achieves state-of-the-art accuracy comparable to full-precision models.

Provides up to 3x speedup on GPU hardware.

Maintains a strict unified NVFP4 format with minimal overhead.

Abstract

The emergence of fine-grained numerical formats like NVFP4 presents new opportunities for efficient Large Language Model (LLM) inference. However, it is difficult to adapt existing Post-Training Quantization (PTQ) strategies to these formats: rotation-based methods compromise fine-grained block isolation; smoothing techniques struggle with significant 4-bit quantization errors; and mixed-precision approaches often conflict with hardware constraints on unified-precision computation. To address these challenges, we propose ARCQuant, a framework that boosts NVFP4 performance via Augmented Residual Channels. Distinct from methods that compromise block isolation or hardware uniformity, ARCQuant maintains a strictly unified NVFP4 format by augmenting the activation matrix with quantized residual channels. This design integrates the error compensation process directly into the matrix reduction dimension, enabling the use of standard, highly optimized GEMM kernels with minimal overhead. Theoretical analysis confirms that the worst-case error bound of our dual-stage NVFP4 quantization is comparable to that of standard 8-bit formats such as MXFP8. Extensive experiments on LLaMA and Qwen models demonstrate that ARCQuant achieves state-of-the-art accuracy, comparable to full-precision baselines in perplexity and downstream tasks. Furthermore, deployment on RTX 5090 and RTX PRO 6000 GPUs confirms practical benefits, achieving up to 3x speedup over FP16. Our code is available at https://github.com/actypedef/ARCQuant .