Saved in:
Bibliographic Details
Main Authors: Chen, Junyu, Li, Jungang, Xiong, Jing, Wang, Wenjie, Yang, Qingyao, Xiao, He, Li, Zhen, Wu, Taiqiang, Chen, Mengzhao, Peng, Zhen, Tao, Chaofan, Shi, Long, Yang, Hongxia, Wong, Ngai
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.04163
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914568419147776
author Chen, Junyu
Li, Jungang
Xiong, Jing
Wang, Wenjie
Yang, Qingyao
Xiao, He
Li, Zhen
Wu, Taiqiang
Chen, Mengzhao
Peng, Zhen
Tao, Chaofan
Shi, Long
Yang, Hongxia
Wong, Ngai
author_facet Chen, Junyu
Li, Jungang
Xiong, Jing
Wang, Wenjie
Yang, Qingyao
Xiao, He
Li, Zhen
Wu, Taiqiang
Chen, Mengzhao
Peng, Zhen
Tao, Chaofan
Shi, Long
Yang, Hongxia
Wong, Ngai
contents Large language model inference is often bounded by memory footprint and bandwidth in resource-constrained deployments, making quantization fundamental to efficient serving. While post-training quantization (PTQ) maintains high fidelity at 4-bit, it deteriorates at 2-3 bits. In essence, existing methods enforce a shape-invariant quantization grid (e.g., the fixed uniform intervals of UINT2) for each group, severely restricting the feasible set for error minimization. To address this, we propose Bit-Plane Decomposition Quantization (BPDQ), which constructs a variable quantization grid via bit-planes and scalar coefficients, and iteratively refines them using second-order information while progressively compensating for quantization errors to minimize output discrepancy. In the 2-bit regime, BPDQ enables serving Qwen2.5-72B on a single RTX 3090 with 83.85\% GSM8K accuracy (vs. 90.83\% at 16-bit). Moreover, we theoretically show that the variable grid expands the feasible set, and that the quantization process consistently aligns with the optimization objective in Hessian-induced geometry. The code is available at https://github.com/KingdalfGoodman/BPDQ.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04163
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models
Chen, Junyu
Li, Jungang
Xiong, Jing
Wang, Wenjie
Yang, Qingyao
Xiao, He
Li, Zhen
Wu, Taiqiang
Chen, Mengzhao
Peng, Zhen
Tao, Chaofan
Shi, Long
Yang, Hongxia
Wong, Ngai
Machine Learning
Large language model inference is often bounded by memory footprint and bandwidth in resource-constrained deployments, making quantization fundamental to efficient serving. While post-training quantization (PTQ) maintains high fidelity at 4-bit, it deteriorates at 2-3 bits. In essence, existing methods enforce a shape-invariant quantization grid (e.g., the fixed uniform intervals of UINT2) for each group, severely restricting the feasible set for error minimization. To address this, we propose Bit-Plane Decomposition Quantization (BPDQ), which constructs a variable quantization grid via bit-planes and scalar coefficients, and iteratively refines them using second-order information while progressively compensating for quantization errors to minimize output discrepancy. In the 2-bit regime, BPDQ enables serving Qwen2.5-72B on a single RTX 3090 with 83.85\% GSM8K accuracy (vs. 90.83\% at 16-bit). Moreover, we theoretically show that the variable grid expands the feasible set, and that the quantization process consistently aligns with the optimization objective in Hessian-induced geometry. The code is available at https://github.com/KingdalfGoodman/BPDQ.
title BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models
topic Machine Learning
url https://arxiv.org/abs/2602.04163