Machine Unlearning on TOFU Italian (1% forget)
32.5Refusal Rate (Forget)Guardrail Classifier
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Guardrail ClassifierInput Variant=Italian2026.05 | 32.5 | 1.5 | |
| ICCUInput Variant=Italian2026.05 | 100 | 8.5 |
Loading the SOTA2 catalog…