X 用户 @_achan96_:8 小时以上任务的部分识别边界窄于短任务,或更易自动化
AI 导读
X 用户 @_achan96_ 观察到,其所讨论方法中 8 小时以上任务的部分识别边界(partial identification bounds)比短任务窄得多;按他的理解,这意味着判断一个 8 小时以上任务是否成功,比判断短任务更容易。他据此提出疑问:8 小时以上任务是否因此更可验证、更易自动化,还是只是方法本身的某种特性。
正文 · 原文
It's pretty interesting that the partial identification bounds for 8+h tasks are way narrower than the bounds for shorter tasks
If I understand the methodology correctly, that means it's way easier to judge whether an 8+h task succeeds than whether a shorter task succeeds
Does that mean that 8+h tasks are more verifiable and hence easier to automate? Or maybe it's just some quirk of the methodology?