BitsMoE: Efficient Spectral Energy-Guided Bit Allocation for MoE LLM Quantization

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhao, Jiayu, Teng, Zihan, Fan, Minhao, Ma, Tianrui, Ren, Wentao, Chen, Song, Liu, Weichen
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916067932110848
author Zhao, Jiayu
Teng, Zihan
Fan, Minhao
Ma, Tianrui
Ren, Wentao
Chen, Song
Liu, Weichen
author_facet Zhao, Jiayu
Teng, Zihan
Fan, Minhao
Ma, Tianrui
Ren, Wentao
Chen, Song
Liu, Weichen
contents Mixture-of-Experts (MoE) large language models reduce per-token computation through sparse expert activation, but their deployment remains memory-intensive because all expert weights must be kept resident in memory. Existing MoE compression methods struggle in the ultra-low-bit regime: pruning irreversibly removes model capacity, while coarse-grained quantization fails to allocate bits according to heterogeneous expert and weight-direction importance. We propose BitsMoE, a spectral-energy-guided bit-allocation framework for MoE LLM quantization. BitsMoE decomposes each MoE layer by SVD into a shared basis and expert-specific spectral factors, retaining the shared basis without quantization to preserve common cross-expert structure and using the expert-specific factors as fine-grained quantization units. To determine the bit-width of each unit, BitsMoE formulates spectrum-wise mixed-precision quantization as an activation-aware reconstruction surrogate and solves an integer linear program that minimizes estimated reconstruction loss under a fixed bit budget. Experiments across multiple MoE LLMs show that BitsMoE substantially reduces downstream task accuracy degradation in ultra-low-bit regimes. Under 2-bit quantization on Qwen3-30B-A3B-Base, BitsMoE accelerates quantization by 12.3$\times$, improves average accuracy by 27.83 percentage points, and increases decoding speed by 1.76$\times$ over GPTQ. Our model and code are publicly available at https://github.com/zjiayu064/BitsMoE.
format Preprint
id arxiv_https___arxiv_org_abs_2606_00079
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle BitsMoE: Efficient Spectral Energy-Guided Bit Allocation for MoE LLM Quantization
Zhao, Jiayu
Teng, Zihan
Fan, Minhao
Ma, Tianrui
Ren, Wentao
Chen, Song
Liu, Weichen
Machine Learning
Artificial Intelligence
Mixture-of-Experts (MoE) large language models reduce per-token computation through sparse expert activation, but their deployment remains memory-intensive because all expert weights must be kept resident in memory. Existing MoE compression methods struggle in the ultra-low-bit regime: pruning irreversibly removes model capacity, while coarse-grained quantization fails to allocate bits according to heterogeneous expert and weight-direction importance. We propose BitsMoE, a spectral-energy-guided bit-allocation framework for MoE LLM quantization. BitsMoE decomposes each MoE layer by SVD into a shared basis and expert-specific spectral factors, retaining the shared basis without quantization to preserve common cross-expert structure and using the expert-specific factors as fine-grained quantization units. To determine the bit-width of each unit, BitsMoE formulates spectrum-wise mixed-precision quantization as an activation-aware reconstruction surrogate and solves an integer linear program that minimizes estimated reconstruction loss under a fixed bit budget. Experiments across multiple MoE LLMs show that BitsMoE substantially reduces downstream task accuracy degradation in ultra-low-bit regimes. Under 2-bit quantization on Qwen3-30B-A3B-Base, BitsMoE accelerates quantization by 12.3$\times$, improves average accuracy by 27.83 percentage points, and increases decoding speed by 1.76$\times$ over GPTQ. Our model and code are publicly available at https://github.com/zjiayu064/BitsMoE.
title BitsMoE: Efficient Spectral Energy-Guided Bit Allocation for MoE LLM Quantization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2606.00079