Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

ConMoE: Expert-Pool Consolidation via Prototype Reassignment for MoE Compression

About

Mixture-of-Experts (MoE) language models reduce per-token computation but still require storing and serving all experts, making deployment memory-intensive. Existing post-training compression methods mainly shrink this cost by pruning experts or merging their weights. We formulate post-training MoE compression as expert-pool consolidation: retaining a smaller set of pretrained experts as reusable prototypes and deterministically remapping each original expert reference to one selected prototype. This view separates the reduced expert pool from the reuse structure that represents the original expert slots, and allows prototype sharing within local layer scopes while preserving the original router interface. We propose ConMoE, a train-free prototype remapping framework that selects retained experts using calibration-based contribution and replaceability signals, then redirects original expert calls to the selected prototypes without weight updates or post-compression fine-tuning. Experiments on three pretrained MoE language models show that ConMoE matches or outperforms strong pruning and merging baselines in several settings, achieving the best average score on deepseek-moe-16b-base at both 25% and 50% routed-expert reduction, while remaining competitive on Qwen3-30B-A3B and OLMoE-1B-7B-0125. Ablations indicate that deterministic reassignment is the most stable component, whereas broader cross-layer sharing and post-hoc weight fusion are model-dependent.

Yilun Yao, Jiaming Pan, Elsie Dai, Peizhuang Cong, Yaoming Li, Tong Yang• 2026

Related benchmarks

TaskDatasetResultRank
Multiple-choice Question AnsweringARC Easy
Accuracy78.6
257
Commonsense Question AnsweringWinoGrande
Accuracy71.5
73
Multiple-choice Question AnsweringPIQA
Accuracy80.4
63
Multiple-choice Question AnsweringBoolQ
MC Accuracy0.885
46
Multiple-choice Question AnsweringHellaSwag
Normalized Accuracy77.2
33
Multiple-choice Question AnsweringARC Challenge
Normalized Accuracy54.6
33
Showing 6 of 6 rows

Other info

Follow for update