AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

MoBE: Mixture-of-Basis-Experts for Compressing MoE-based LLMs

Xiaodong Chen, Mingming Ha, Zhenzhong Lan, Jing Zhang, Jianguo Li

ICLR 2026 · Architectures and efficiency › Mixture of experts · approach: Efficient MoE language model systems

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
use MoBE instead of MoLAE for compressionQwen3-30B-A3B-2507 · 15-benchmark averageaccuracy (%): 67.5 -> 75.8Improved
use MoBE instead of D2-MoEQwen3-30B-A3B-2507 · 15-benchmark averageaccuracy (%): 66.4 -> 75.8Improved
use MoBE instead of MoLAEKimi-K2-Instruct · 15-benchmark averageaccuracy (%): 76.6 -> 81.1Improved
use MoBE instead of MoLAEDeepSeek-V3-0324 · 15-benchmark averageaccuracy (%): 73.1 -> 78.0Improved
use MoBE instead of MoLAEQwen3-235B-A22B-2507 · 15-benchmark averageaccuracy (%): 73.2 -> 80.9Improved
compress with MoBE vs uncompressedLing-Lite-Chat · 15-benchmark averageaccuracy (%): 59.7 -> 58.6Hurt
reduce active experts (MoBE-dagger) vs MoBE (ablation)Qwen3-235B-A22B-2507 · 15-benchmark averageaccuracy (%): 80.9 -> 79.7Hurt
reduce active experts (MoBE-dagger) vs MoBE (ablation)Qwen3-30B-A3B-2507 · 15-benchmark averageaccuracy (%): 75.8 -> 75.7No clear effect
use ReLU activation instead of SiLU/Tanh in basis combination (ablation)MoBE (Qwen3-30B gate matrices) · per-layer reconstruction MSEMSEHurt
use Sigmoid instead of no activation (ablation)MoBE (Qwen3-30B gate matrices) · per-layer reconstruction MSEMSEHurt
add SiLU/Tanh/GELU activation vs none (ablation)MoBE (Qwen3-30B gate matrices) · per-layer reconstruction MSEMSEImproved
add Z-score normalization of expert weights (ablation)MoBE (Qwen3-30B gate matrices) · per-layer reconstruction MSEMSEImproved

Later papers that compared with this method

2 later accepted papers report MoBE as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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