AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

Parameters as Experts: Adapting Vision Models with Dynamic Parameter Routing

Meng Lou, Stanley Yu, Yizhou Yu

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: Mixture-of-experts LoRA · 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
share one expert center across all 18 layers of a stage instead of groups of 3 layers (ablation)ParaX · COCO2017 detection, Swin-BAPb: 46.3 -> 47.3Improved
share expert center across 9 layers instead of 18 (ablation)ParaX · COCO2017 detection, Swin-BAPb: 47.1 -> 47.3Improved
activate all experts instead of sparse top-K (ablation)ParaX · COCO2017 detection, Swin-BAPb: K=3 46.3 -> all 47.4Improved
use softmax instead of ReLU in the router (ablation)ParaX · COCO2017 detection, Swin-BAPb: 45.8 -> 47.3Improved
use softmax instead of Sigmoid in the router (ablation)ParaX · COCO2017 detection, Swin-BAPb: 46.2 -> 47.3Improved
use multi-scale kernels [3,5,7] instead of a single kernel (ablation)ParaX · COCO2017 detection, Swin-BAPb: 46.3 -> 47.1Improved
add spatially-varying aggregation (SA) module (ablation)ParaX · COCO2017 detection, Swin-BAPb: 47.1 -> 47.3Improved
increase kernel sizes to [5,7,9] (ablation)ParaX · COCO2017 detection, Swin-BAPb: 47.1 -> 46.9No clear effect
add residual connections to sequential multi-kernel convs (ablation)ParaX · COCO2017 detection, Swin-BAPb: 46.6 -> 47.3Improved
change weight init (kaiming normal / uniform vs trunc normal) (ablation)ParaX · COCO2017 detection, Swin-BAPb: 47.3 vs 47.1 / 47.2No clear effect
trade latent dim for expert capacity (M=L, C=128 vs M=4L, C=40) (ablation)ParaX · COCO2017 detection, Swin-BAPb: 46.1 -> 47.3Improved
increase router hidden size from 16 to 40 (ablation)ParaX · COCO2017 detection, Swin-BAPb: 47.3 -> 47.4No clear effect

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