494
Apple study exposes deep cracks in LLMs’ “reasoning” capabilities
(arstechnica.com)
This is a most excellent place for technology news and articles.
LLMs don’t see words, they see tokens. They were always just guessing