News for hackers. Stuff that matters.
Re:OpenLLaMA: An Open Reproduction of LLaMA by visarga on Wednesday May 03, 2023 @04:52PM (#35807637)
Oh no, it will hallucinate an obscure fact, but not basics. It's pretty good at reciting theory, it would pass many ML engineering theoretical interviews.

If you don't trust its memory, copy a piece of high quality text in the topic of interest inside the context, as reference.

Re:OpenLLaMA: An Open Reproduction of LLaMA by agentdrtran on Monday May 08, 2023 @04:21PM (#35866553)
it's repeatedly made up entire quotes and research papers?