Safety Gate Capacity Scaling on Safety Scaling Testbed N=100-10,000
12.4Classifier Ceiling (B=0.01)Classifier-based safety gate
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Classifier-based safety gateN=10,0002026.03 | 12.4 | — | — | |
| Classifier-based safety gateN=1,0002026.03 | 3.9 | — | — | |
| Classifier-based safety gateN=1002026.03 | 1.2 | — | — | |
| Ball verifierN=1002026.03 | — | 50 | 42 | |
| Ball verifierN=1,0002026.03 | — | 500 | 128 | |
| Ball verifierN=10,0002026.03 | — | 5,000 | 403 |