LLM training itself is unsupervised

LLM (no semantics) + fine-tune model which usually is supervised to get some fancy application.

I think coding is an example. It can "solve" LeetCode which all learned from the existing solutions.

所有跟帖： 

• language? I see. Thanks. 我相關的不在那個方向 -成功的兔- ♀ (0 bytes) () 08/18/2023 postreply 19:34:50