ScaleBITS: Scalable Bitwidth Search for Hardware-Aligned Mixed-Precision LLMs

TL;DR

ScaleBITS introduces a scalable, hardware-efficient mixed-precision quantization method for large language models, optimizing bitwidth allocation to significantly reduce memory and inference costs while maintaining performance.

Contribution

It presents a novel sensitivity analysis, hardware-aligned weight partitioning, and a scalable optimization algorithm for automated mixed-precision quantization.

Findings

Up to 36% improvement over uniform quantization

Outperforms state-of-the-art sensitivity-aware methods by up to 13%

Achieves ultra-low-bit quantization without runtime overhead

Abstract

Post-training weight quantization is crucial for reducing the memory and inference cost of large language models (LLMs), yet pushing the average precision below 4 bits remains challenging due to highly non-uniform weight sensitivity and the lack of principled precision allocation. Existing solutions use irregular fine-grained mixed-precision with high runtime overhead or rely on heuristics or highly constrained precision allocation strategies. In this work, we propose ScaleBITS, a mixed-precision quantization framework that enables automated, fine-grained bitwidth allocation under a memory budget while preserving hardware efficiency. Guided by a new sensitivity analysis, we introduce a hardware-aligned, block-wise weight partitioning scheme, powered by bi-directional channel reordering. We formulate global bitwidth allocation as a constrained optimization problem and develop a scalable approximation to the greedy algorithm, enabling end-to-end principled allocation. Experiments show that ScaleBITS significantly improves over uniform-precision quantization (up to +36%) and outperforms state-of-the-art sensitivity-aware baselines (up to +13%) in ultra-low-bit regime, without adding runtime overhead.