作者位次 #2/3ICML 2026Poster· Social Aspects / Accountability, Transparency, and Interpretability
The Expert Strikes Back: Interpreting Mixture-of-Experts Language Models at Expert Level
Jeremy Herbst · Stefan Wermter · Jae Hee Lee
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Jeremy Herbst · Stefan Wermter · Jae Hee Lee
Jiaye Lin · Mengdi Li · Xufeng Zhao · Wenhao Lu · Peilin Zhao · Stefan Wermter · Di Wang
Reward models trained through Reinforcement Learning from AI Feedback (RLAIF) methods frequently suffer from limited generalizability, which hinders the alignment performance of policy models. This challenge stems from various issues, including distribution shift, preference label noise, and mism...