SHAPE applies coalition-aware Shapley values to prune experts in MoE LLMs, retaining competitive accuracy at 20-40% pruning rates on Qwen3-30B-A3B, GPT-OSS-20B, and DeepSeek-V2-Lite without retraining.
Seap: Training-free sparse expert activation pruning unlock the brainpower of large language models,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SHAPE: Coalition-Aware Expert Pruning for Sparse Mixture-of-Experts LLMs
SHAPE applies coalition-aware Shapley values to prune experts in MoE LLMs, retaining competitive accuracy at 20-40% pruning rates on Qwen3-30B-A3B, GPT-OSS-20B, and DeepSeek-V2-Lite without retraining.