LLM Unlearning Should Be Form-Independent

Open in new window