Anthropic@AnthropicAI新Anthropic Fellows研究：模型规范中期训练（MSM）。 标准的对齐方法通过期望行为的示例来训练AI。但这可能无法泛化到新情境。 MSM通过首先教导AI我们希望它们如何泛化以及原因，来解决这一问题。
New Anthropic Fellows research: Model Spec Midtraining (MSM).
Standard alignment methods train AIs on examples of desired behavior. But this can fail to generalize to new situations.
MSM addresses this by first teaching AIs how we would like them to generalize and why.
来源：Anthropic · x.com
新Anthropic Fellows研究：模型规范中期训练（MSM）。 标准的对齐方法通过期望行为的示例来训练AI。但这可能无法泛化到新情境。 MSM通过首先教导AI我们希望它们如何泛化以及原因，来解决这一问题。
New Anthropic Fellows research: Model Spec Midtraining (MSM).
Standard alignment methods train AIs on examples of desired behavior. But this can fail to generalize to new situations.
MSM addresses this by first teaching AIs how we would like them to generalize and why.
来源：Anthropic· x.com