Weak-to-strong generalization

T1Covered by 1 source · first reported by OpenAI News at 14 Dec 2023, 00:00 UTC

We present a new research direction for superalignment, together with promising initial results: can we leverage the generalization properties of deep learning to control strong models with weak supervisors?

Sources

• T1Weak-to-strong generalizationOpenAI News14 Dec 2023, 00:00