Safety Decision Making on 230-task benchmark
92.8AR-SSafeGate
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SafeGate2026.04 | 92.8 | 0 | 9.2 | 4.8 | 4 | 28.3 | |
| SELP2026.04 | 85.5 | 26 | — | — | — | — | |
| RoboGuard2026.04 | 74.7 | 39 | — | — | — | — | |
| GPT-4o2026.04 | 51.8 | 0 | 40.6 | 48.2 | 11 | 91.3 | |
| Gemini 2.5-flash2026.04 | 12 | 0 | 51.5 | 79.5 | 19 | 71.7 |