AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

MASQuant: Modality-Aware Smoothing Quantization for Multimodal Large Language Models

Lulu Hu, Wenhu Xiao, Xin Chen, Xinhua Xu, Bowen Xu, Kun Li, Yongliang Tao

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multimodal fusion with missing modalities · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add modality-aware smoothing (MAS) instead of uniform smoothing (ablation)MASQuant W4A8 · Librispeech (Omni-3B)WER: 77.4 -> 3.8Improved
add modality-aware smoothing with learned optimization (ablation)MASQuant W4A8 · OmniBench (Omni-3B)accuracy: 33.3 -> 47.7Improved
add modality-aware smoothing with learned optimization (ablation)MASQuant W4A8 · MMMU (VL-3B)accuracy: 33.3 -> 46.7Improved
use equal modality loss weights (1.0/1.0) vs lowering vision weight to 0.5 (ablation)MASQuant W4A8 · MMMU (VL-3B)accuracy: 30.0 -> 37.8Improved
use equal modality loss weights vs lambda_v=0.1 (ablation)MASQuant W4A8 · PPL (VL-3B)PPL: 33.7 -> 17.2Improved
train smoothing factors 2 epochs instead of 1 (ablation)MASQuant W4A8 · average accuracy (VL-3B)average accuracy: 56.9 -> 61.2Improved
train more than 2 epochs (5 or 10) (ablation)MASQuant W4A8 · average accuracy (VL-3B)average accuracy: 61.2 -> 59.2 (5 epochs), 60.8 (10 epochs)No clear effect
apply MASQuant vs MBQ at W4A8 on VL-3BQwen2.5-VL-3B · multimodal averageaverage accuracy: 64.6 -> 64.7No clear effect
apply MASQuant vs MBQ at W4A8 on VL-7BQwen2.5-VL-7B · multimodal averageaverage accuracy: 68.5 -> 69.0Improved
apply MASQuant vs MBQ at W8A8 on VL-3BQwen2.5-VL-3B · multimodal averageaverage accuracy: 70.1 -> 70.9Improved
apply MASQuant vs MBQ on omni audio at W4A8Qwen2.5-Omni-3B · LibrispeechWER: 9.5 -> 3.6Improved
apply MASQuant vs MBQ on Wenetspeech at W4A8Qwen2.5-Omni-3B · WenetspeechWER: 8.5 -> 8.7Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy