作者位次 #3/6ICML 2026Poster· Social Aspects / Safety
Sparse Models, Sparse Safety: Unsafe Routes in Mixture-of-Experts LLMs
Yukun Jiang · Hai Huang · Mingjie Li · Yage Zhang · Michael Backes · Yang Zhang
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。
Yukun Jiang · Hai Huang · Mingjie Li · Yage Zhang · Michael Backes · Yang Zhang
Wai Man Si · Mingjie Li · Michael Backes · Yang Zhang
Machine learning models are increasingly deployed in real-world applications, but even aligned models such as Mistral and LLaVA still exhibit unsafe behaviors inherited from pre-training. Current alignment methods like SFT and RLHF primarily encourage models to generate preferred responses, but d...
Mingjie Li · Wai Man Si · Michael Backes · Yang Zhang
As high-quality data becomes harder to obtain, reward models are increasingly important. Beyond the costly RLHF stage, they are now used at inference time to guide LLM generation and in data selection for post-training. These methods bring efficiency and performance gains, but current reward mode...