MODE: Modality-Decomposed Expert-Level Mixed-Precision Quantization for MoE Multimodal LLMs
Mixture-of-Experts Multimodal Large Language Models (MoE-MLLMs) offer remarkable performance but incur high GPU memory costs, making compression essential. Existing expert-level mixed-precision quantization degrades on MoE-MLLMs due to two biases: vision tokens dominate cross-modal expert selection, and redundant tokens skew intra-vision frequency. MODE addresses this by decomposing selection frequency by modality, filtering redundant tokens, and evaluating per-modality quantization sensitivity, using integer linear programming to assign bit-widths per expert. At W3A16, average performance loss is limited to 2.9%.
[2606.17118] MODE: Modality-Decomposed Expert-Level Mixed-Precision Quantization for MoE Multimodal LLMs
[Submitted on 15 Jun 2026]
Title:MODE: Modality-Decomposed Expert-Level Mixed-Precision Quantization for MoE Multimodal LLMs
View a PDF of the paper titled MODE: Modality-Decomposed Expert-Level Mixed-Precision Quantization for MoE Multimodal LLMs, by Yuanteng Chen and 11 other authors
View PDF HTML (experimental)
Abstract:Mixture-of-Experts Multimodal Large Language Models (MoE-MLLMs) offer remarkable performance but incur prohibitive GPU memory costs, making compression essential. Among PTQ methods, expert-level mixed-precision quantization has proven effective for MoE-LLMs, yet suffers notable degradation on MoE-MLLMs due to two overlooked biases in expert importance estimation. (1) At the cross-modal level, the numerical dominance of vision tokens causes expert selection frequency to be dominated by vision tokens, masking experts that are critical to the text modality; (2) at the intra-vision level, the large proportion of redundant vision tokens further skew frequency statistics, obscuring experts critical for informative visual content. To bridge gaps, we propose MODE, a modality-decomposed expert-level mixed-precision quantization framework for MoE-MLLMs that decomposes expert selection frequency by modality, filters redundant vision tokens to obtain denoised visual frequency, and further evaluates quantization sensitivity per modality as a complementary signal to frequency-based estimation. These signals are integrated into an Integer Linear Programming formulation to assign per-expert bit-widths under a given budget. Extensive experiments show that MODE is particularly well-suited for MoE-MLLMs, limiting average performance loss to within 2.9% at W3A16, with larger gains at the extreme 2-bit setting.
Comments: 18 pages, 8 figures
Subjects:
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2606.17118 [cs.LG]
(or arXiv:2606.17118v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.17118
arXiv-issued DOI via DataCite
Submission history
From: Yuanteng Chen [view email] [v1] Mon, 15 Jun 2026 10:59:11 UTC (1,302 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled MODE: Modality-Decomposed Expert-Level Mixed-Precision Quantization for MoE Multimodal LLMs, by Yuanteng Chen and 11 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.LG
new | recent | 2026-06
Change to browse by:
cs cs.AI
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
IArxiv recommender toggle
IArxiv Recommender (What is IArxiv?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)