AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Deconstructing Pre-training: Knowledge Attribution Analysis in MoE and Dense Models

Bo Wang, Junzhuo Li, Hong Chen, Yuanlin Chu, Yuxuan Fan, Xuming Hu

AAAI 2026 · Architectures and efficiency › Mixture of experts · approach: Expert routing and specialization · 3 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
use MoE (OLMoE) instead of dense (OLMo) architecturepre-training · FFN neuronstop-1% set stability / Rt: Rt 26.4 -> 48.9Improved
use MoE instead of dense architecturepre-training · FFN layers (10K-600K)layer-distribution consistency rho_avg: 0.54 -> 0.97Improved
use MoE instead of dense architecturepre-training · ATTN layers (10K-600K)sigma_rel: 8.64 -> 0.49Improved
mask top-1% FFN neurons (ablation)OLMoE vs OLMo · relational HIT@10performance drop: OLMoE 35.47% vs OLMo 96.19% dropHurt
mask top-10 attention heads (ablation)OLMoE vs OLMo · relational HIT@10performance drop: OLMoE 9.44% vs OLMo 50.43% dropHurt
mask the single most important attention head (ablation)OLMoE vs OLMo · relational HIT@10performance drop: OLMoE 0.06% vs OLMo 16.46% dropMixed

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