News for hackers. Stuff that matters.
Re:Can Phi3 and Llama3 Generalise to Bio-Medical Tasks? An E... by DrRavenstein on Friday April 26, 2024 @09:00AM (#40168906)
I've been investigating how well the smallest Phi3 and Llama3 models do at BioASQ - a biomedical Q&A reading comprehension task.

Llama 3 8B is certainly quite competitive and compelling. Phi doesn't do quite as well but there may be opportunities for fine tuning.