AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Breaking the Echo Chamber: A Dynamic Ensemble Pruning Perspective on MoE

Xinlai Kang, Dunyao Xue, Zhengbo Wang, Chengshuo Du, Xinghao Chen, Hang Zhou, Hanting Chen, Cheng Meng

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: Expert routing and specialization

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
add Mahalanobis diversity objective instead of l1-norm criterion (Sigma = identity) (ablation)MP-MoE on OLMoE-1B-7B · OLMoE-1B-7B ablationbenchmark performance: -3.00%Improved
use normalized covariance estimator instead of raw co-occurrence matrix C (ablation)MP-MoE on OLMoE-1B-7B · OLMoE-1B-7B ablationbenchmark performance: -1.37%Improved
add MP warmup schedule (standard routing for first 1% of steps) (ablation)MP-MoE on OLMoE-1B-7B · OLMoE-1B-7B ablationbenchmark performance: -0.55%Improved
replace top-k routing with MP-MoE routingOLMoE-1B-7B, 1B tokens · Six-benchmark averageaverage score: 0.282 -> 0.312Improved
replace top-k routing with MP-MoE routingOLMoE-1B-7B, 5B tokens · Six-benchmark averageaverage score: 0.294 -> 0.319Improved
replace top-k routing with MP-MoE routingOLMoE-0.2B-1B, 50B tokens · Six-benchmark averageaverage score: 0.279 -> 0.291Improved
replace top-k routing with MP-MoE routingOLMoE-0.2B-1B, 50B tokens · HellaSwagaccuracy: 0.246 -> 0.243No clear effect
replace top-k routing with MP-MoE routingOLMoE-0.2B-1B, 50B tokens · ARC-Challengeaccuracy: 0.255 -> 0.257No clear effect
replace top-k routing with MP-MoE routingOLMoE-1B-7B, 1B tokens · BBHexact match: 0.121 -> 0.169Improved
replace top-k routing with MP-MoE routingOLMoE-1B-7B, 1B tokens · BoolQaccuracy: 0.563 -> 0.597Improved
replace top-k routing with MP-MoE routingFlame-MoE (38M active parameters) · pre-training perplexityPPL: 4.71% reduction in PPLImproved
replace top-k routing with SimSMoE (competing diversity method)Flame-MoE (38M active parameters) · pre-training perplexityPPL: 1.31% reductionImproved

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