Papers12:02Chinese contentGricean Retreat:探针揭示LLM知道边界但不愿撤退这篇用探针方法拆开了LLM一本正经胡说八道的原因:模型内部其实知道自己不知道,但嘴硬不肯说泛话。#T-REx#LLM#knowledge boundaries#AI safetyaarXiv cs.AI@Dananjay Srinivas et al. (3)SourceRead laterReadWorth followingUsefulFollow T-REx