Jailbreak Attack on RoboTwin LingBot-VA
1.6Misclassification Failure Rate (MFR)clean
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| cleanEvaluation Source=Human2026.04 | 1.6 | 0 | 1.6 | |
| cleanEvaluation Source=RD2026.04 | 2.2 | 0 | 2.2 | |
| TPAEvaluation Source=Human2026.04 | 4.2 | 0 | 4.2 | |
| RSAEvaluation Source=Human2026.04 | 5.2 | 0 | 5.2 | |
| TPAEvaluation Source=RD2026.04 | 6.4 | 0 | 6.4 | |
| RSAEvaluation Source=RD2026.04 | 7.8 | 0 | 7.8 | |
| JailWAMEvaluation Source=Human2026.04 | 62 | 22.2 | 84.2 | |
| JailWAMEvaluation Source=RD2026.04 | 69.4 | 17.6 | 87 |