works on
deep learning 1inference divergence 1numerical precision 1reduction order 1sparse mixture-of-experts 1

From the 1 of 1 linked paper with an AI index.

collaborators

1 paper

cs.LG2026

From Expert Reduction to Behavioral Divergence: Tracing Numerical State through Sparse MoE Inference

Tianyang Zhu

The paper studies how different expert reduction orders and numerical precision settings cause divergent behavior in sparse mixture-of-experts (MoE) inference, showing that identic…