Machine Unlearning on TOFU Italian (5% forget split)
23.5Refusal Rate (Forget)Guardrail Classifier
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Guardrail ClassifierInput Variant=Italian2026.05 | 23.5 | 0.3 | |
| ICCUInput Variant=Italian2026.05 | 96 | 5.1 |
Loading the SOTA2 catalog…