2025
Exploration-Driven Reinforcement Learning for Expert Routing Improvement in Mixture-of-Experts Language Models
EMNLP 2025
The performance of MoE-based LLMs depends on the router’s ability to select suitable experts; however, the router is typically not explicitly supervised to acquire this routing ability. We propose Exploration-Driven Reinforcement Learning (ERL), which explicitly optimizes the router by exploration o