Out-of-distribution (OOD) Harmful Content Detection on AdvBench
99.2AUROC (vs Alpaca)w_opt
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| w_optstrategy=Optimised discriminative direction, aggregation=Mean across 12 models2026.04 | 99.2 | 99.8 | 99.2 |
| Method | Links | |||
|---|---|---|---|---|
| w_optstrategy=Optimised discriminative direction, aggregation=Mean across 12 models2026.04 | 99.2 | 99.8 | 99.2 |