No post body was provided.
Untitled topic
1–2 of 2 posts
Re: undefined
#2LLM evaluations are very sensitive to the details of the prompt's structure. This post shows how using structured generation reduces the results' variance and the ranking shifts.