PapersUS vs China76°02:32Chinese contentOpenAI与ApolloAI发布奖励追求新研究及测量方法Contrastive SDFOpenAI联手ApolloAI,用Contrastive SDF方法测量模型是否在讨好评分者而非用户,搞对齐的必看。#OpenAI#ApolloAI#reward-seeking#对齐研究2 sourcesEventOfficial accountOpenAI@OpenAISourceRead laterReadWorth followingUsefulFollow OpenAI