AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Efficient and Effective Weight-Ensembling Mixture of Experts for Multi-Task Model Merging

Li Shen, Anke Tang, Enneng Yang, Guibing Guo, Yong Luo, Lefei Zhang, Xiaochun Cao, Bo Du, Dacheng Tao

TPAMI 2025 · Architectures and efficiency › Mixture of experts · approach: Mixture-of-experts LoRA · 28 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 WEMoE dynamic merging instead of static AdaMerging (layer)CLIP-ViT-B/32 · 8 tasks avgaverage accuracy (%): 80.9 -> 89.4Improved
use WEMoE instead of traditional MTLCLIP-ViT-B/32 · 8 tasks avgaverage accuracy (%): 88.9 -> 89.4Improved
prune MLP task vectors to 90% sparsity and share routing across layers (E-WEMoE-90%)WEMoE · 8 tasks avg (ViT-L/14)average accuracy (%): 93.6 -> 93.5No clear effect
increase task vector sparsity from 90% to 99% (ablation)E-WEMoE · 8 tasks avg (ViT-B/32)average accuracy (%): 89.1 -> 86.8Hurt
add 2 routing layers instead of 0 (ablation)WEMoE · 6 seen tasks avg (ViT-B/32)average accuracy (%): 73.5 -> 86.5Improved
add 2 routing layers instead of 0 (ablation)WEMoE · 2 unseen tasks avg (ViT-B/32)average accuracy (%): 64.2 -> 71.1Improved
use WEMoE instead of AdaMerging on unseen tasksCLIP-ViT-B/32 · 2 unseen tasks avgaverage accuracy (%): 70.0 -> 71.1Improved
use E-WEMoE instead of WEMoE on unseen tasksWEMoE · 2 unseen tasks avg (ViT-B/32)average accuracy (%): 71.1 -> 61.7Hurt
use WEMoE instead of AdaMerging under motion blurCLIP-ViT-B/32 · Motion blur corrupted testaverage accuracy (%): 80.6 -> 87.6Improved
use WEMoE under pixelate corruptionCLIP-ViT-B/32 · Pixelate corrupted testaverage accuracy (%): 42.3 -> 27.3Hurt
upcycle entire transformer block instead of MLP only (ablation)WEMoE · Up-scaling strategy, 8 tasksaverage accuracy (%): 88.3 -> 86.4Hurt
upcycle attention and MLP separately instead of MLP only (ablation)WEMoE · Up-scaling strategy, 8 tasksaverage accuracy (%): 88.3 -> 89.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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