Cohere Labs - Chuanyang Zheng and Jiankai Sun
other
other
Date: Dec 19, 2025
Time: 6:00 PM - 7:00 PM
Location: Online
Learning more about MoE routing as Nadaraya–Watson regression and proposes a zero-cost FFN-style kernel-inspired router (KERN) that generalizes softmax/sigmoid routing and performs better using GELU + normalization from first authors!
Chuanyang Zheng is a Machine Learning Researcher at Morgan Stanley and Jiankai Sun is a PhD candidate at Stanford University.
Add event to calendar