Beyond Benchmarks: Understanding Mixture-of-Experts Models through Internal Mechanisms
2509.23933v1
cs.LG, cs.CL
2025-10-01
Авторы:
Jiahao Ying, Mingbao Lin, Qianru Sun, Yixin Cao
Abstract
Mixture-of-Experts (MoE) architectures have emerged as a promising direction,
offering efficiency and scalability by activating only a subset of parameters
during inference. However, current research remains largely
performance-centric, with limited understanding of its internal mechanisms,
thereby constraining broader progress. In this work, we use an internal metric
to investigate the mechanisms of MoE architecture by explicitly incorporating
routing mechanisms and analyzing expert-level behaviors. Through systematic
analyses of a wide range of publicly available MoE models, we uncover several
findings: (1) neuron utilization decreases as models evolve, reflecting
stronger generalization; (2) training exhibits a dynamic trajectory, where
benchmark performance alone provides limited signal while MUI reveals deeper
insights; (3) task completion emerges from collaborative contributions of
multiple experts, with shared experts driving concentration; and (4) activation
patterns at the neuron level provide a fine-grained proxy for data diversity.
Together, these results demonstrate the potential of MUI as a complementary
indicator to benchmark performance, offering new insights into the capacity,
dynamics, and specialization of MoE models. Our project can be found at
https://yingjiahao14.github.io/MoE-MUI/.
Ссылки и действия
Дополнительные ресурсы: