paper-with-me

Papers

Is Retraining-Free Enough? The Necessity of Router Calibration for Efficient MoE Compression

2026-02-10 · Sieun Hyeon, Jaeyoung Do arxiv

Mixture-of-Experts (MoE) models scale capacity efficiently, but their massive parameter footprint creates a deployment-time memory bottleneck. We organize retraining-free MoE compression into three paradigms - Expert Pruning, Expert Editing, and Expert Merging - and show that persistent post-compression degradation largely stems from a neglected factor: router-expert mismatch when experts are changed but the router is left untouched. We argue that effective retraining-free compression should avoid updating expert parameters while allowing lightweight router calibration. To this end, we propose Router Knowledge Distillation (Router KD), which updates only a tiny fraction of parameters (the router) by distilling the original model's next-token distribution on unlabeled calibration data. Experiments across representative methods in all three paradigms demonstrate consistent performance recovery, with substantially larger gains in fine-grained MoEs (many small experts) than in coarse-grained MoEs due to their more complex routing decision boundaries.

📄 PDF Abstract BibTeX arXiv:2603.02217

Code (0)

등록된 구현이 없습니다.

Tasks

Knowledge Distillation

Similar Papers 제목 키워드 기반

When Model Merging Breaks Routing: Training-Free Calibration for MoE

2026-06-02 · Canbin Huang, Tianyuan Shi, Xiaojun Quan, Jingang Wang 외 arxiv

Model merging has emerged as a cost-effective approach for consolidating the capabilities of multiple LLMs without retraining. However, existing merging techniques, largely based on linear parameter arithmetic or optimiz…

Mathematical ReasoningCode Generation

SCOPE-Router: Cost-Aware Open-Set VLM Routing for Execution-Oriented Tasks

2026-08-12 · Tao Yu, Yifei Qu, Zhiqing Cui, Pengfei Zhou 외 arxiv

Model routing aims to select the most suitable model from a candidate pool for each query, balancing quality and cost. Existing VLM routing research is limited to traditional VQA evaluation, lacks systematic calibration …

ACE: Adaptive Calibration-Free Expert Skipping for MoE-based LLMs

2026-09-04 · Zukang Xu, Zhixiong Zhao, Xing Hu, Jiangyong Yu 외 arxiv

Mixture-of-Experts (MoE) architectures provide an efficient paradigm for scaling large language models (LLMs), yet fixed top-k routing activates the same number of expert slots for every token, causing substantial redund…

ProxRouter: Proximity-Weighted LLM Query Routing for Improved Robustness to Outliers

2025-10-10 · Shivam Patel, Neharika Jali, Ankur Mallick, Gauri Joshi arxiv

Large language model (LLM) query routers are critical to modern AI platforms as they seek to improve efficiency by assigning inference queries to accurate, yet low-cost models. Parametric routers typically use trained ne…

PreMoE: Proactive Inference for Efficient Mixture-of-Experts

2025-05-23 · Zehua Pei, Ying Zhang, Hui-Ling Zhen, Tao Yuan 외 arxiv

Mixture-of-Experts (MoE) models offer dynamic computation, but are typically deployed as static full-capacity models, missing opportunities for deployment-specific specialization. We introduce PreMoE, a training-free fra…