In a 26-layer MoE model, injecting Gaussian weight errors into middle-layer experts hurts math accuracy most, while deep-layer errors can sometimes improve instruction compliance.
In2023 IEEE/ACM International Conference on Computer Aided Design (ICCAD)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
MoE-Compression: How the Compression Error of Experts Affects the Inference Accuracy of MoE Model?
In a 26-layer MoE model, injecting Gaussian weight errors into middle-layer experts hurts math accuracy most, while deep-layer errors can sometimes improve instruction compliance.