News for hackers. Stuff that matters.

A comment on We Built a Language Model 14,000,000x Smaller Than GPT3 and Formally Verified It. Parent comment

Re:We Built a Language Model 14,000,000x Smaller Than GPT3 a... by icodar on Tuesday July 22, 2025 @07:52AM (#44645733)
The next token prediction appears to be predicted based on fixed grammatical rules. However, modern LLMs learn the rules themselves. Did I misunderstand?
[ Parent | on HN ]

There are no comments on this comment yet.