AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

LD-MoLE: Learnable Dynamic Routing for Mixture of LoRA Experts

Yuan Zhuang, Yi Shen, Yuexin Bian, Qing Su, Shihao Ji, Yuanyuan Shi, Fei Miao

ICLR 2026 · Architectures and efficiency › Mixture of experts · approach: Efficient MoE language model systems · 8 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 TopK routing with learnable dynamic Sparsegen routingMoLA(8888) · 9 benchmarks, Qwen3-1.7Baverage score: 79.24 -> 81.63Improved
replace TopK routing with learnable dynamic Sparsegen routingMoLA(8888) · 9 benchmarks, Llama3.2-3Baverage score: 78.18 -> 82.05Improved
replace ReLU routing with Sparsegen routing and predicted lambdaReMoLE · 9 benchmarks, Llama3.1-8Baverage score: 84.47 -> 85.18Improved
replace ReLU routing with Sparsegen routing and predicted lambdaReMoLE · 9 benchmarks, Qwen3-1.7Baverage score: 81.02 -> 81.63Improved
prune experts per layer (2468) instead of uniform 8MoLA · 9 benchmarks, Llama3.2-3Baverage score: 78.18 -> 78.72Improved
use predicted lambda instead of fixed lambda 0.5 (ablation)LD-MoLE · 5 datasets, Qwen3-1.7Baverage score: 83.66 -> 84.56Improved
use predicted lambda instead of fixed lambda -10.0 (ablation)LD-MoLE · 5 datasets, Qwen3-1.7Baverage score: 83.35 -> 84.56Improved
add sparsity loss with coefficient 1.0 (ablation)LD-MoLE · 5 datasets, Qwen3-1.7Baverage score: 84.56 -> 83.11Hurt
add sparsity loss with coefficient 0.01 (ablation)LD-MoLE · 5 datasets, Qwen3-1.7Baverage score: 84.56 -> 84.02Hurt
add sparsity loss with coefficient 0.01 (beta) (ablation)LD-MoLE · Qwen3-1.7BMFLOPs: 106 -> 100Improved
add sparsity loss with beta 1.0 (ablation)LD-MoLE · Qwen3-1.7BMFLOPs: 106 -> 83Improved
use shared MLP instead of local linear lambda predictor (ablation)LD-MoLE · 8 tasks, Qwen3-1.7Baverage score: 84.65 -> 85.05Improved

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