|
|
|
|
|
|
by ACCount37
26 days ago
|
|
|
For something like "a hard limit" to hold, LLMs must be restricted to only reproducing existing text. This is utterly false even for base models - their basin seems to be "permutations loosely inspired by existing text". And that's before all the post-training comes in. What's the "limit" there? |
|
|