Post

Can anyone explain why LLMs keep needing more data to feed on? What's the incremental benefit of a new text?