Prompt Injection Detection on SCOUT-450 (Held-out evaluation)
92.4AccuracyGPT-4o
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4oLat (ms)=14572026.05 | 92.4 | 11.8 | 2.1 | |
| GPT-5.1Lat (ms)=14642026.05 | 87.3 | 16.9 | 7.2 | |
| DeepSeek-V4Lat (ms)=92242026.05 | 87.1 | 21.6 | 1.5 | |
| GPT-5.2Lat (ms)=36382026.05 | 80.4 | 5.1 | 38.5 | |
| Gemini-3.1Lat (ms)=48062026.05 | 66.9 | 55.7 | 3.6 |