MoDES：通过动态专家跳过加速混合专家多模态大语言模型
MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping
MoDES 提供了一种免训练的专家跳过框架，可直接用于加速基于 MoE 的多模态大语言模型（如 Qwen3-VL-MoE）的视觉语言推理，通过全局调制局部门控和双模态阈值化，在跳过 88% 专家时仍保持甚至提升精度，并显著减少预填充和解码时间，为视觉研究者部署高效 MLLM 提供了可复用的技术手段。
原文依据 MoDES, the first training-free framework that adaptively skips experts to enable efficient and accurate MoE MLLM inference.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论混合专家（MoE）多模态大语言模型（MLLMs）在视觉语言任务上表现出色，但存在计算效率低下的问题。为减少推理开销，已有研究提出基于当前输入令牌停用冗余专家的专家跳过方法。然而，我们发现将这些原本为单模态大语言模型（LLMs）设计的方法应用于 MLLMs 会导致显著的性能下降。这主要是因为此类方法未能考虑专家在 MoE 层间的异质贡献以及层内令牌的模态特定行为。受这些发现启发，我们提出了 MoDES，这是首个免训练框架，能够自适应地跳过专家以实现高效准确的 MoE MLLM 推理。它包含一种全局调制局部门控（GMLG）机制，将全局层级重要性融入局部路由概率，以准确估计每个令牌的专家重要性。随后应用双模态阈值化（DMT）方法，分别处理来自每个模态的令牌，以推导跳过调度。为设置最优阈值，我们引入了一种利用单调性性质的前沿搜索算法，将收敛时间从数天缩短至数小时。在 13 个基准上对 3 个模型系列的广泛实验表明，MoDES 远超先前方法。例如，当对 Qwen3-VL-MoE-30B-A3B-Instruct 跳过 88% 的专家时，性能提升高达 10.67%（97.33% 对 86.66%）。此外，MoDES 显著提升了推理速度，预填充时间提升 2.16 倍，解码时间提升 1.26 倍。
英文原文摘要 1716 字符
Mixture-of-Experts (MoE) Multimodal large language models (MLLMs) excel at vision-language tasks, but they suffer from high computational inefficiency. To reduce inference overhead, expert skipping methods have been proposed to deactivate redundant experts based on the current input tokens. However, we find that applying these methods--originally designed for unimodal large language models (LLMs)--to MLLMs results in considerable performance degradation. This is primarily because such methods fail to account for the heterogeneous contributions of experts across MoE layers and modality-specific behaviors of tokens within these layers. Motivated by these findings, we propose MoDES, the first training-free framework that adaptively skips experts to enable efficient and accurate MoE MLLM inference. It incorporates a globally-modulated local gating (GMLG) mechanism that integrates global layer-wise importance into local routing probabilities to accurately estimate per-token expert importance. A dual-modality thresholding (DMT) method is then applied, which processes tokens from each modality separately, to derive the skipping schedule. To set the optimal thresholds, we introduce a frontier search algorithm that exploits monotonicity properties, cutting convergence time from several days to a few hours. Extensive experiments for 3 model series across 13 benchmarks demonstrate that MoDES far outperforms previous approaches. For instance, when skipping 88% experts for Qwen3-VL-MoE-30B-A3B-Instruct, the performance boost is up to **10.67%** (97.33% vs. 86.66%). Furthermore, MoDES significantly enhances inference speed, improving the prefilling time by **2.16x** and the decoding time by **1.26x**.