AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

GeMoE: Gating Entropy is All You Need for Uncertainty-aware Adaptive Routing in MoE-based Large Vision-Language Models

Chaoxiang Cai, Minghe Weng, Jie Li, Yibo Jiang, Longrong Yang, Zequn Qin, Xi Li

ECCV 2026 · Architectures and efficiency › Mixture of experts · approach: Efficient MoE language model systems · 0 citations (OpenAlex)

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 gating-entropy guidance via monotonic loss (ablation)GeMoE · MMB/SQA/GQAaverage performance: 60.03 -> 64.77Improved
enforce positive instead of negative entropy-k correlation (ablation)GeMoE · MMB/SQA/GQAaverage performance: 51.57 -> 64.77Improved
set L_mono weight alpha from 0.0 to 1.0 (ablation)GeMoE · MolmoE four-benchmark averageaverage performance: 65.79 -> 71.74Improved
increase alpha from 1.0 to 1.5 (ablation)GeMoE · MolmoE four-benchmark averageaverage performance: 71.74 -> 71.27No clear effect
replace static Top-8 with GeMoE adaptive routingMolmoE-1B-7B · Six-benchmark averageaverage performance: 63.99 -> 63.61No clear effect
replace static Top-6 with GeMoE adaptive routingDeepSeek-VL2-Tiny · Six-benchmark averageaverage performance: 71.29 -> 70.90No clear effect
replace static Top-8 with GeMoEMolmoE-1B-7B · Inference FLOPsGFLOPs per token: 80.65 -> 68.45Improved
replace static Top-8 with GeMoEMolmoE-1B-7B · Throughputtokens/s: 1709 -> 1828Improved
replace AdaMoE with GeMoEMolmoE-1B-7B · Six-benchmark averageaverage performance: 61.77 -> 63.61Improved
replace Top-p dynamic routing with GeMoEMolmoE-1B-7B · Six-benchmark averageaverage performance: 61.18 -> 63.61Improved
reduce static Top-k from 8 to 5MolmoE-1B-7B · Six-benchmark averageaverage performance: 63.99 -> 60.07Hurt
replace Top-5 static routing with GeMoE (5.43 experts)MolmoE-1B-7B · Six-benchmark averageaverage performance: 60.07 -> 63.61Improved

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