AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

GEMQ: Global Expert-Level Mixed-Precision Quantization for MoE LLMs

Jianing Deng, Song Wang, Dongwei Wang, Zijie Liu, Tianlong Chen, Huanrui Yang, Jingtong Hu

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: MoE LLM compression · 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
replace layer-wise local bit allocation (PMQ) with global expert-level LP allocation plus router fine-tuningMixtral-8x7B 1.5 bits · C4 perplexityperplexity: 20.77 -> 16.20Improved
add global router fine-tuning after quantization (ablation)GEMQ Mixtral-8x7B 1.5 bits · WikiText2 / C4perplexity: WT2 9.66 -> 7.69; C4 23.53 -> 17.18Improved
replace quantized router logits with recorded FP logits (rigid FP-matching) (ablation)GEMQ Mixtral-8x7B 1.5 bits · WikiText2 / C4perplexity: WT2 9.66 -> 9.29; C4 23.53 -> 22.72No clear effect
estimate importance with a 2-bit quantized model instead of FP (progressive quantization) (ablation)GEMQ Mixtral-8x7B 1.5 bits · WikiText2perplexity: 9.66 -> 9.33Improved
fine-tune routers of the 2-bit model before using it for importance estimation (ablation)GEMQ Mixtral-8x7B 1.5 bits · WikiText2perplexity: 9.33 -> 9.01Improved
calibrate on C4 only vs C4 only for GSM8KGEMQ vs PMQ Mixtral-8x7B 2.5 bits · GSM8Kaccuracy (%): PMQ 33.06 vs GEMQ 31.77Hurt
add 128 MATH sequences to C4 calibration set (ablation)GEMQ Mixtral-8x7B 2.5 bits · GSM8Kaccuracy (%): 31.77 -> 42.30Improved
naive PMQ extension: apply layer-wise coefficients in a global LP (ablation)global expert bit allocation Mixtral-8x7B · C4 perplexityperplexityHurt
apply FIM approximation to weight Hessian instead of output Hessian (ablation)global expert bit allocation Mixtral-8x7B · C4 perplexityperplexityHurt
two-step allocation (layer budgets then experts) vs single global LP (ablation)global expert bit allocation Mixtral-8x7B · C4 perplexityperplexityHurt
increase calibration sequences beyond 128 (ablation)GEMQ · calibration sizequantization performanceNo clear effect
GEMQ vs SpQR at 1.5 bitsMixtral-8x7B · 7-task zero-shot averageaccuracy (%): 31.87 -> 52.00Improved

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