23 ms·
Probably the only way to do it reliably would be to intercept the prompt with a specially trained classifier? I think you're right that once it gets to the main
by danenania 1y ago
Probably the only way to do it reliably would be to intercept the prompt with a specially trained classifier? I think you're right that once it gets to the main model, nothing really works.