Machine Unlearning on TOFU Chinese (5% forget)
10Refusal Rate (Forget)Guardrail Classifier
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Guardrail ClassifierInput Variant=Chinese2026.05 | 10 | 0.7 | |
| ICCUInput Variant=Chinese2026.05 | 96 | 30 |
Loading the SOTA2 catalog…
| Method | Links | ||
|---|---|---|---|
| Guardrail ClassifierInput Variant=Chinese2026.05 | 10 | 0.7 | |
| ICCUInput Variant=Chinese2026.05 | 96 | 30 |