Pre-load security auditing on SkillGuardBench Main
99.26Overall Exact AccuracySKILLGUARD-ROBUST
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SKILLGUARD-ROBUST2026.04 | 99.26 | 100 | 99.22 | 100 | 98.87 | 100 | 100 | |
| SKILLGUARD-CALIBRATE2026.04 | 98 | 100 | 97.61 | 100 | 97.27 | 100 | 99.66 | |
| SKILLGUARD-VERIFY2026.04 | 96.75 | 100 | 95.31 | 100 | 95.52 | 100 | 99.26 | |
| Qwen2.5-14BPackage Mode=full-package2026.04 | 89.53 | 97.25 | 87.66 | 85.85 | 88.37 | 86.86 | 98.17 | |
| PEBPackage Mode=full-package2026.04 | 82.78 | 92.01 | 80.44 | 78.62 | 80.96 | 79.42 | 95.38 |