Physical Red Teaming on RedVLA Cumulative-Level Scenarios
100Attack Success Rate (ASR)OpenVLA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenVLAScenario=Dangerous Item Misuse2026.04 | 100 | 70 | |
| OpenVLA-OFTScenario=Dangerous Item Misuse2026.04 | 100 | 36.7 | |
| VLA-AdapterScenario=Dangerous Item Misuse2026.04 | 100 | 86.7 | |
| VLA-Adapter-ProScenario=Dangerous Item Misuse2026.04 | 100 | 43.3 | |
| π0Scenario=Dangerous Item Misuse2026.04 | 100 | 86.7 | |
| π0.5Scenario=Dangerous Item Misuse2026.04 | 100 | 86.7 | |
| π0.5Scenario=Resource Damage2026.04 | 98.5 | 76.9 | |
| OpenVLAScenario=Resource Damage2026.04 | 97.7 | 62.3 | |
| π0Scenario=Resource Damage2026.04 | 97.7 | 74.6 | |
| VLA-Adapter-ProScenario=Resource Damage2026.04 | 96.9 | 70 | |
| π0Scenario=Robot Damage2026.04 | 96.7 | 15 | |
| OpenVLA-OFTScenario=Resource Damage2026.04 | 95.4 | 56.9 | |
| VLA-AdapterScenario=Resource Damage2026.04 | 95.4 | 65.4 | |
| VLA-AdapterScenario=Robot Damage2026.04 | 95 | 13.3 | |
| π0.5Scenario=Robot Damage2026.04 | 91.7 | 36.7 | |
| VLA-Adapter-ProScenario=Robot Damage2026.04 | 88.3 | 3.3 | |
| OpenVLA-OFTScenario=Robot Damage2026.04 | 70 | 11.7 | |
| OpenVLAScenario=Robot Damage2026.04 | 61.7 | 5 |