AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

CommitMoE: Efficient Fallback-Free MoE Inference with Offloading Under GPU Memory Constraints

Han Li, Jingwei Sun, Junqing Lin, Guangzhong Sun

AAAI 2026 · Architectures and efficiency › Mixture of experts · approach: Efficient MoE language model systems · 1 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 mispredicted experts with predicted ones when no predicted expert is correct (ablation)Qwen1.5-MoE-Chat · GSM8Kaccuracy: 53.22% -> 54.51%No clear effect
replace mispredicted experts with predicted ones when no predicted expert is correct (ablation)DeepSeek-V2-Lite-Chat · BBHaccuracy: 49.12% -> 49.16%No clear effect
add Output Weight Adjustment with tuned alpha1, alpha2 (ablation)CommitMoE · DeepSeek-V2-Lite-Chat averageaverage score: 48.16 -> 48.48Improved
add Output Weight Adjustment with tuned alpha1, alpha2 (ablation)CommitMoE · Qwen1.5-MoE-Chat averageaverage score: 42.38 -> 43.06Improved
add Output Weight Adjustment with tuned alpha1, alpha2 (ablation)CommitMoE · Mixtral-8x7B-Instruct averageaverage score: 51.02 -> 50.73Hurt
use tuned model-specific OWA hyperparameters instead of neutral alpha1 = alpha2 = 1.0 (ablation)CommitMoE OWA · Qwen1.5-MoE-Chat averageaverage score: 42.52 -> 43.06Improved
use tuned model-specific OWA hyperparameters instead of neutral alpha1 = alpha2 = 1.0 (ablation)CommitMoE OWA · Mixtral-8x7B-Instruct averageaverage score: 50.45 -> 50.73Improved
fallback-free CommitMoE versus prefetching with fallback (+Prefetch using Commit Router) (ablation)CommitMoE · end-to-end latency, Environment 1speedupImproved
CommitMoE versus vLLM on slower GPU and PCIeCommitMoE · Environment 2 (RTX 2080Ti, PCIe Gen3)speedup: up to 4.3x Mixtral, 9.4x Qwen, 7.5x DeepSeekImproved
CommitMoE versus vLLM on faster hardwareCommitMoE · Environment 1 (RTX 4090)speedup: 1.3-1.5x Mixtral; up to 1.8x Qwen/DeepSeekImproved

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