logoalt Hacker News

vatsachak • yesterday at 3:11 PM • 1 reply • view on HN

If none of the findings overlapped then LLMs are either

- For some reason choosing only a subset of issues.

- Different LLMs actually learn different things.

I suspect the former since they do this for me and it's extremely annoying sometimes


Replies

qup • yesterday at 5:11 PM

3 - non deterministic, lazy