Hacker News new | ask | show | jobs
by ctxc 416 days ago
The error introduced by the data is expected and internalized, it's the error of LLMs on _top_ of that that's hard to.