3 ms·
Why should they be? The token is at the word level more or less, so the model can't see the letters to substitute. When we get character-level models this shou
by GardenLetter27 2y ago
Why should they be? The token is at the word level more or less, so the model can't see the letters to substitute.
When we get character-level models this should be more viable.