Ban&Pick: Ehancing Performance and Efficiency of MoE-LLMs via Smarter Routing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Yuanteng, Wang, Peisong, Shao, Yuantian, Zeng, Nanxin, Xu, Chang, Cheng, Jian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912616701493248
author Chen, Yuanteng
Wang, Peisong
Shao, Yuantian
Zeng, Nanxin
Xu, Chang
Cheng, Jian
author_facet Chen, Yuanteng
Wang, Peisong
Shao, Yuantian
Zeng, Nanxin
Xu, Chang
Cheng, Jian
contents Sparse Mixture-of-Experts (MoE) has become a key architecture for scaling large language models (LLMs) efficiently. Recent fine-grained MoE designs introduce hundreds of experts per layer, with multiple experts activated per token, enabling stronger specialization. However, during pre-training, routers are optimized mainly for stability and robustness: they converge prematurely and enforce balanced usage, limiting the full potential of model performance and efficiency at inference. In this work, we uncover two overlooked issues: (i) a few highly influential experts are underutilized due to premature and balanced routing decisions; and (ii) enforcing a fixed number of active experts per token introduces substantial redundancy. Instead of retraining models or redesigning MoE architectures, we introduce Ban&Pick, a post-training, plug-and-play strategy for smarter routing. Pick discovers and reinforces key experts-a small group with outsized impact on performance-leading to notable accuracy gains across domains. Ban further dynamically prunes redundant experts based on layer and token sensitivity, delivering faster inference with minimal accuracy loss. Experiments on fine-grained MoE-LLMs (DeepSeek, Qwen3) across math, code, and general reasoning benchmarks demonstrate that Ban\&Pick delivers free performance gains and inference acceleration without retraining or architectural changes. For instance, on Qwen3-30B-A3B, it improves accuracy from 80.67 to 84.66 on AIME2024 and from 65.66 to 68.18 on GPQA-Diamond, while accelerating inference by 1.25x under the vLLM.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06346
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Ban&Pick: Ehancing Performance and Efficiency of MoE-LLMs via Smarter Routing
Chen, Yuanteng
Wang, Peisong
Shao, Yuantian
Zeng, Nanxin
Xu, Chang
Cheng, Jian
Machine Learning
Artificial Intelligence
Sparse Mixture-of-Experts (MoE) has become a key architecture for scaling large language models (LLMs) efficiently. Recent fine-grained MoE designs introduce hundreds of experts per layer, with multiple experts activated per token, enabling stronger specialization. However, during pre-training, routers are optimized mainly for stability and robustness: they converge prematurely and enforce balanced usage, limiting the full potential of model performance and efficiency at inference. In this work, we uncover two overlooked issues: (i) a few highly influential experts are underutilized due to premature and balanced routing decisions; and (ii) enforcing a fixed number of active experts per token introduces substantial redundancy. Instead of retraining models or redesigning MoE architectures, we introduce Ban&Pick, a post-training, plug-and-play strategy for smarter routing. Pick discovers and reinforces key experts-a small group with outsized impact on performance-leading to notable accuracy gains across domains. Ban further dynamically prunes redundant experts based on layer and token sensitivity, delivering faster inference with minimal accuracy loss. Experiments on fine-grained MoE-LLMs (DeepSeek, Qwen3) across math, code, and general reasoning benchmarks demonstrate that Ban\&Pick delivers free performance gains and inference acceleration without retraining or architectural changes. For instance, on Qwen3-30B-A3B, it improves accuracy from 80.67 to 84.66 on AIME2024 and from 65.66 to 68.18 on GPQA-Diamond, while accelerating inference by 1.25x under the vLLM.
title Ban&Pick: Ehancing Performance and Efficiency of MoE-LLMs via Smarter Routing
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2509.06346