Safety Detection on BeaverTails (held-out)
76.7AUROCMultiLayer-Linear
Evaluation Results
| Method | Links | |
|---|---|---|
| MultiLayer-LinearProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 76.7 | |
| Geometry-LiteProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 76.3 | |
| Best single-layer probeProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 75.4 | |
| TaT-Disp-LSTMProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 74.2 | |
| MultiLayer-DIMProtocol=Leave-one-benchmark-out (LOBO), Aggregation=Mean across nine model backbones2026.05 | 73.3 |