gitaskhub

Locating and editing refusal in the J-space workspace with the Jacobian lens: refusal is legible ~10 layers before the first token, and only ~1/3 lives in the verbalizable workspace.

Language · Python
License · Apache-2.0
Ask anything about this repo to start.
Full explanation on explaingit →

By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).