Safety detection on TS-Bench AgentDojo-Traj (eval)
1.36Efficiency (s/sample)TS-Guard
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TS-Guard2026.01 | 1.36 | 86.18 | |
| GPT-4o2026.01 | 1.98 | 56.59 | |
| AGrailModel=GPT-4o2026.01 | 8.75 | 58.05 | |
| ShieldAgent2026.01 | 10 | — |
| Method | Links | ||
|---|---|---|---|
| TS-Guard2026.01 | 1.36 | 86.18 | |
| GPT-4o2026.01 | 1.98 | 56.59 | |
| AGrailModel=GPT-4o2026.01 | 8.75 | 58.05 | |
| ShieldAgent2026.01 | 10 | — |