AI News

OpenAI 与 ApolloResearch、METR 等四家机构开展嵌入评估

OpenAI更新于 11月17日 00:13

We’re working with @ApolloResearch, @METR_Evals, @TransluceAI, and @GoodfireAI on embedded assessments: giving outside experts deep access to our internal model

AI 导读

OpenAI 宣布与 ApolloResearch、METR_Evals、TransluceAI 和 GoodfireAI 合作开展嵌入评估,让外部专家深入访问其内部模型和系统,以调查对齐问题并测试安全防护措施。ApolloResearch 将评估模型是否会暗中违背开发者意图以及防护措施能否发现;METR_Evals 将调查其迄今最大规模强化学习运行中的失准行为,并已初步同意在就范围达成一致后进行更广泛的监控评估。TransluceAI 将评估后训练对齐评测的质量与覆盖范围并计划在 2027 年扩大试点,GoodfireAI 则派驻可解释性研究人员用其平台分析模型内部,聚焦训练运行监控、调试与对齐。

推荐理由

AI 评分 62

OpenAI 把内部模型与系统开放给四家外部机构做嵌入评估,读者可了解各家分别查什么、边界在哪。

主题

标签

正文

We’re working with @ApolloResearch, @METR_Evals, @TransluceAI, and @GoodfireAI on embedded assessments: giving outside experts deep access to our internal models and systems to investigate alignment and test our safeguards.

These assessments build on our ongoing alignment and safety research, bringing independent scrutiny to how we identify and manage risks in our systems.

https://t.co/SNAAG0uHmp

来源

OpenAI