← 研踪 PaperTrailarXiv:2611.20885 · v1
ProSE: Probing Reconstruction Error for Semantic-guided Expert Skipping in MoE-MLLMs
首次提交 2026-11-16 18:56
版本更新 2026-11-16 18:56
雷达观测 2026-11-17 17:11
摘要
Mixture-of-Experts Multimodal Large Language Models (MoE-MLLMs) offer scalable inference through sparse expert activation, yet substantial computational redundancy remains among the activated experts. Existing expert skipping methods typically calibrate thresholds based on target skipping ratios or final-output divergence, overlooking the accumulation of skipping-induced error across layers and the varying semantic importance of individual tokens. To address that, we propose ProSE, a training-free, semantic-guided expert skipping framework that probes layer-wise reconstruction error to promote representation consistency compared with default activation. Specifically, we construct query-aware semantic masks to identify salient visual tokens and progressively calibrate modality-specific skipping thresholds for each MoE layer. By minimizing masked reconstruction error under a given budget, ProSE maintains semantic consistency after skipping and allows adaptive expert allocation under limited activation budgets. Extensive experiments on 9 benchmarks demonstrate consistent improvements on average over existing expert skipping methods, particularly under aggressive skipping settings. ProSE maintains over 96\% performance on KimiVL-A3B-Inst with 83\% expert skipping and accelerates prefilling by .
作者备注:8 pages
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-16 18:56 · 最新观测版本
ProSE: Probing Reconstruction Error for Semantic-guided Expert Skipping in MoE-MLLMs
Mixture-of-Experts Multimodal Large Language Models (MoE-MLLMs) offer scalable inference through sparse expert activation, yet substantial computational redundancy remains among the activated experts. Existing expert skipping methods typically calibrate thresholds based on target skipping ratios or final-output divergence, overlooking the accumulation of skipping-induced error across layers and the varying semantic importance of individual tokens. To address that, we propose ProSE, a training-free, semantic-guided expert skipping framework that probes layer-wise reconstruction error to promote representation consistency compared with default activation. Specifically, we construct query-aware semantic masks to identify salient visual tokens and progressively calibrate modality-specific skipping thresholds for each MoE layer. By minimizing masked reconstruction error under a given budget, ProSE maintains semantic consistency after skipping and allows adaptive expert allocation under limited activation budgets. Extensive experiments on 9 benchmarks demonstrate consistent improvements on average over existing expert skipping methods, particularly under aggressive skipping settings. ProSE maintains over 96\% performance on KimiVL-A3B-Inst with 83\% expert skipping and accelerates prefilling by .
查看此版本解读 →查看此版本原文 ↗