Exploration-Driven Reinforcement Learning for Expert Routing Improvement in Mixture-of-Experts Language Models.

Gyunyeop Kim, Sangwoo Kang

Browse the full EMNLP paper archive.