List your company

Machine Unlearning on TOFU (Retain99)

3.2Refusal RateGuardrail Filter
2.6126.58110.5514.519May 26, 2026

Evaluation results

MethodLinks
Guardrail FilterBackbone=Llama-2-7B-chat-hf2026.053.2
ICCUBackbone=Llama-3-8B-Instruct2026.057.5
ICCUBackbone=Llama-2-7B-chat-hf2026.058
Guardrail ClassifierBackbone=Llama-3-8B-Instruct2026.059.7
Guardrail ClassifierBackbone=Llama-2-7B-chat-hf2026.059.7
Guardrail FilterBackbone=Llama-3-8B-Instruct2026.0517.9