AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

MetaMoE: Diversity-Aware Proxy Selection for Privacy-Preserving Mixture-of-Experts Unification

Weisen Jiang, Shuhao Chen, Sinno Jialin Pan

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 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 relevance-weighted DPP proxy selection (ablation)MetaMoE · LLaMA-3.1-8B NLP averageaccuracy: 80.32 -> 81.59Improved
add relevance-weighted DPP proxy selection (ablation)MetaMoE · CLIP ViT-B/16 CV averageaccuracy: 95.39 -> 96.24Improved
add relevance-weighted DPP proxy selection (ablation)FlexOlmo · LLaMA-3.1-8B NLP averageaccuracy: 77.46 -> 79.78Improved
add relevance-weighted DPP proxy selection (ablation)FlexOlmo · LLaMA-3.2-3B NLP averageaccuracy: 72.50 -> 73.35Improved
add context-aware router (sequence-level context) (ablation)MetaMoE · LLaMA-3.2-3B NLP averageaccuracy: 72.62 -> 74.42Improved
add context-aware router (sequence-level context) (ablation)MetaMoE · CLIP ViT-B/16 CV averageaccuracy: 95.94 -> 96.24Improved
add proxy-aligned expert training (ablation)MetaMoE · CLIP ViT-B/16 CV averageaccuracy: 94.09 -> 96.24Improved
add proxy-aligned expert training (ablation)MetaMoE · LLaMA-3.1-8B NLP averageaccuracy: 80.99 -> 81.59Improved
train experts solely on proxy data instead of private plus proxy data (ablation)MetaMoE · CLIP ViT-B/16 CVaccuracy: substantial accuracy degradation across all domainsHurt
use relevance-weighted DPP proxies with context-aware router and proxy-aligned training vs FlexOlmoMetaMoE · CLIP ViT-B/32 CV averageaccuracy: 92.92 -> 94.52Improved
swap public dataset Alpaca for OpenOrcaMetaMoE · LLaMA-3.2-3B NLP averageaccuracy: 74.42 -> 74.79No clear effect
MetaMoE vs unrestricted MoE trained on pooled private dataMetaMoE · LLaMA-3.1-8B NLP averageaccuracy: 81.34 -> 81.59Improved

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