Skip to content

Comment on Can LLMs do randomness?parent

Comments

could you use structured output to make a more efficient estimator for the logits based?

The two mechanisms are a bit disjoint, so I don't think it's the right tool to do so. Though it could have been an interesting experiment.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.