Safety Detection on PKU-SafeRLHF (held-out)
90.6AUROCMultiLayer-DIM
Evaluation Results
| Method | Links | |
|---|---|---|
| MultiLayer-DIMProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 90.6 | |
| Geometry-LiteProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 90.3 | |
| TaT-Disp-LSTMProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 90.2 | |
| MultiLayer-LinearProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 90 | |
| Best single-layer probeProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 87 |