Physical Red Teaming on RedVLA State-Level Scenarios
98.4Attack Success Rate (ASR)π0
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| π0Scenario=Robot Damage2026.04 | 98.4 | 75 | |
| π0.5Scenario=Robot Damage2026.04 | 98.4 | 80 | |
| VLA-AdapterScenario=Resource Damage2026.04 | 97.1 | 60 | |
| OpenVLAScenario=Resource Damage2026.04 | 96.9 | 47.7 | |
| VLA-Adapter-ProScenario=Robot Damage2026.04 | 96.9 | 60 | |
| π0Scenario=Resource Damage2026.04 | 96.6 | 73.1 | |
| OpenVLAScenario=Robot Damage2026.04 | 96.6 | 53.4 | |
| OpenVLA-OFTScenario=Resource Damage2026.04 | 96.3 | 70.9 | |
| VLA-Adapter-ProScenario=Resource Damage2026.04 | 96.3 | 67.4 | |
| π0.5Scenario=Resource Damage2026.04 | 96.3 | 85.1 | |
| VLA-AdapterScenario=Robot Damage2026.04 | 96.2 | 56.2 | |
| π0.5Scenario=Dangerous Item Misuse2026.04 | 95 | 80.6 | |
| π0Scenario=Dangerous Item Misuse2026.04 | 94.7 | 75.6 | |
| OpenVLA-OFTScenario=Robot Damage2026.04 | 94.7 | 74.7 | |
| VLA-Adapter-ProScenario=Dangerous Item Misuse2026.04 | 94.1 | 60.9 | |
| OpenVLA-OFTScenario=Dangerous Item Misuse2026.04 | 93.8 | 75.6 | |
| OpenVLAScenario=Dangerous Item Misuse2026.04 | 91.2 | 60.9 | |
| VLA-AdapterScenario=Dangerous Item Misuse2026.04 | 88.8 | 61.3 |