AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Synergistic Intra- and Cross-Layer Regularization Losses for MoE Expert Specialization

Rizhen Hu, Yuan Cao, Boao Kong, Mou Sun, Kun Yuan

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: Expert routing and specialization · 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 intra-layer specialization and cross-layer coupling lossesvanilla MoE small · C4 validationperplexity: 14.01 -> 13.75Improved
add sp and cp losses on top of load balance and z-lossDeepSeek-style MoE medium · C4 validationperplexity: 12.07 -> 11.99Improved
add orthogonality and logit-variance auxiliary loss (Guo et al.)load-balance MoE · C4 validationperplexity: 14.01 -> 14.27Hurt
add sp and cp lossesDeepSeek-MoE-16B LoRA SFT · MMLUaccuracy: 0.4143 -> 0.4586Improved
add sp and cp lossesDeepSeek-V2-Lite LoRA SFT · MMLU-Proaccuracy: 0.2605 -> 0.3108Improved
add sp and cp lossesLing-mini-2.0 with AuxLossFree · GSM8Kaccuracy: 0.8173 -> 0.8559Improved
add sp and cp lossesQwen3-30B-A3B full-parameter FT · HumanEvalpass@1: 92.07 -> 95.73Improved
add z-loss to load balance (ablation)DeepSeek-MoE-16B LoRA SFT · MMLUaccuracy: 0.4143 -> 0.3717Hurt
add sp and cp losses; fewer activated expertsmedium MoE · C4 validationperplexity: 12.13 (N=6) vs 12.15 (baseline N=10)Improved
add sp and cp losses; smaller expert poolmedium MoE · C4 validationperplexity: 11.93 (E=96) vs 11.94 (baseline E=112)Improved
add specialization loss to load balance (ablation)1.1B MoE pretraining · training perplexityperplexityImproved
add sp loss to model with cp loss (ablation)0.4B MoE · coupling lossRcpImproved

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