AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

SCHUR-A*: Layer-wise Optimal Expert Pruning for MoEs via Schur-Complement Guided A* Search

Zheng Chen, Yang Weifeng, Jianxiao Tang, Buhui Yao

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: Expert routing and specialization

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
use original gate weights (Mask-Only) instead of refitted coefficients at inference (ablation)SCHUR-A* on Qwen3-30B-A3B · PPLperplexity: 8.84 -> 8.25Improved
use original gate weights (Mask-Only) instead of refitted coefficients at inference (ablation)SCHUR-A* on ERNIE-4.5-21B-A3B · PPLperplexity: 8.12 -> 7.55Improved
use original gate weights (Mask-Only) instead of refitted coefficients at inference (ablation)SCHUR-A* on Qwen3-30B-A3B · MMLUaccuracy: 0.579 -> 0.623Improved
use original gate weights (Mask-Only) instead of refitted coefficients at inference (ablation)SCHUR-A* on Qwen3-30B-A3B · 8-task averageaverage accuracy: 0.658 -> 0.661Improved
add A* global search instead of greedy selection with same gain metric (ablation)SCHUR-A* on Qwen3-30B-A3B · PPLperplexity: 10.30 -> 8.25Improved
add A* global search instead of greedy selection with same gain metric (ablation)SCHUR-A* on Qwen3-30B-A3B · MMLUaccuracy: 0.580 -> 0.623Improved
add A* global search instead of greedy selection with same gain metric (ablation)SCHUR-A* on Qwen3-30B-A3B · ARC-caccuracy: 0.449 -> 0.491Improved
add A* global search instead of greedy selection with same gain metric (ablation)SCHUR-A* on Qwen3-30B-A3B · OpenBookQAaccuracy: 0.432 -> 0.452Improved
add adaptive Kneedle layer-wise budgeting instead of fixed 50% pruning (ablation)SCHUR-A* on Qwen3-30B-A3B-Instruct-2507 · MMLUaccuracy: 0.550 -> 0.672Improved
add adaptive Kneedle layer-wise budgeting instead of fixed 50% pruning (ablation)SCHUR-A* on Qwen3-30B-A3B-Instruct-2507 · GSM8Kaccuracy: +19.6%Improved
use SCHUR-A* criterion instead of REAP at the same fixed 42% pruning ratio (ablation)expert pruning on Qwen3-30B-A3B · 8-task averageaverage accuracy: 0.645 -> 0.661Improved
apply Kneedle adaptive budget to REAP baseline (ablation)REAP on Qwen3-30B-A3B · 8-task averageaverage accuracy: 0.645 -> 0.642No 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