Preprint
Aug 2026
Benchmarking Composable Compression Techniques in Mixture-of-Experts LLMs
The results reveal non-trivial interactions among compression methods: composable compression cannot be predicted from standalone techniques, compression rate alone does not reliably predict quality loss or runtime gain, expert pruning is the dominant degradation source, and average quality can hide workload and architecture-specific failures.
A. Benazir, Chen Chen, Rongxiao Qu et al.
· 0 citations