Coarse Unified Face Attack Detection on UniAttackData Protocol 2
100Real Face ACCGLM-4.6V
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GLM-4.6V2026.05 | 100 | 0 | 0 | 100 | 0.43 | 49.8 | 6.7 | 56.2 | |
| Gemini-3-Pro2026.05 | 100 | 0 | 72.8 | 26.1 | 14.6 | 42.9 | 31.4 | 36.9 | |
| UniShieldParameters=7B2026.05 | 99.3 | 0.67 | 99.8 | 0.22 | 99.4 | 0.37 | 99.5 | 0.36 | |
| UniShieldParameters=3B2026.05 | 99 | 1 | 97.2 | 1.4 | 98.9 | 1.09 | 98.6 | 1.14 | |
| HiPTune2026.05 | 96.1 | 1.96 | 99.3 | 0.33 | 98.7 | 0.67 | 98.9 | 0.69 | |
| CDCN2026.05 | 89 | 11 | 97.3 | 2.67 | 98.9 | 1.11 | 98 | 2.04 | |
| UniAttackDetection2026.05 | 88 | 6 | 99.1 | 0.44 | 99.6 | 0.19 | 98.8 | 0.61 | |
| CLIP2026.05 | 86.7 | 6.67 | 99.1 | 0.44 | 99.6 | 0.23 | 98.7 | 0.68 | |
| CoOp2026.05 | 75.7 | 12.2 | 99.7 | 0.17 | 98.4 | 0.79 | 97.2 | 1.39 | |
| GPT-5.22026.05 | 38.3 | 30.8 | 1.67 | 49.2 | 67 | 27 | 53 | 31.5 |