Prompt Injection Detection on IHEval Rule-following
0.01FPR
AlignSentinel (Enc-first)
Updated 4d ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
2026.02 | 0.01 | 0.1 | |
2026.02 | 0.02 | 0.16 | |
2026.02 | 0.03 | 0.04 | |
2026.02 | 0.06 | 0.09 | |
2026.02 | 0.07 | 0.1 | |
2026.02 | 0.08 | 0.14 |
AlignSentinel (Enc-first)
| Method | Links | ||
|---|---|---|---|
2026.02 | 0.01 | 0.1 | |
2026.02 | 0.02 | 0.16 | |
2026.02 | 0.03 | 0.04 | |
2026.02 | 0.06 | 0.09 | |
2026.02 | 0.07 | 0.1 | |
2026.02 | 0.08 | 0.14 |