AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Expert Merging in Sparse Mixture of Experts with Nash Bargaining

Dung Viet Nguyen, Anh Nguyen Thi, Minh Hoang Nguyen, Luc Nguyen, Shiqi Jiang, Ethan Fetaya, Linh Duy Tran, Gal Chechik, Tan Minh Nguyen

ICLR 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
replace EP-CAMEx's propagation with Nash bargaining weights (NAMEx)EP-CAMEx · WikiText-103 small testperplexity: 85.03 -> 84.12Improved
add Nash bargaining to EP-CAMExEP-CAMEx · ImageNet-1k Acc@1accuracy: 83.23 -> 84.06Improved
add Nash bargaining at ImageNet-A without momentumEP-CAMEx · ImageNet-Aaccuracy: 24.22 -> 25.32Improved
add complex momentum to expert propagationNAMEx · ImageNet-Aaccuracy: 25.32 -> 35.05Improved
add complex momentumNAMEx-Full · WikiText-103 small testperplexity: 83.16 -> 82.94Improved
use real momentum (phi=0) instead of complex momentum (ablation)EP-CAMEx · GLUE RTEaccuracy: 76.23 -> 72.20Hurt
solve bargaining per-layer more often (delta l 1 or 5) vs once (ablation)NAMEx · GLUE RTEaccuracy: 75.09 -> 77.98 (runtime 0.69 -> 1.14 s)Improved
replace complex momentum with quaternion momentum (ablation)NAMEx-Mom · GLUE MRPCaccuracy: 93.02 -> 93.24Improved
use mean instead of zero as disagreement point (ablation)NAMEx-Full · MMLU/GSM8K/ARC DeepSeek-MoEaccuracy: deltas about 0.04No clear effect
increase bargaining iterations from 2 to 60 (ablation)NAMEx-Full · Qwen1.5-MoE and DeepSeek-MoEMMLU/GSM8K/ARC: e.g. MMLU 61.87 -> 61.81No clear effect
add NAMEx-Full to DeepSeek-MoE on GSM8KDeepSeek-MoE · GSM8K zero-shot (Linear router)accuracy: 16.53 -> 16.77Improved
add NAMEx to SMoE for CoLA/SST-2 classification variantsSMoE Top-2 · GLUE SST-2accuracy: 94.35 vs NAMEx 94.46Mixed

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