In Mixture-of-Experts language models, the router typically selects and weights experts based on the token's hidden state, utilizing limited contextual infor…
机构:Amazon
来源:arXiv 2609.20974 | AI4Papers 论文推荐平台