List your company

Button Pressing on VLA Backdoor Evaluation Tasks

0.984Success RateNone
0.800960.848480.8960.94352Jan 7, 2026

Evaluation results

MethodLinks
NoneVLA Model=π02026.010.984—
State BackdoorVLA Model=π02026.010.9640.953
BadVLAVLA Model=π02026.010.9310.708
State BackdoorVLA Model=OpenVLA2026.010.9210.946
NoneVLA Model=OpenVLA2026.010.92—
TrojanRobotVLA Model=π02026.010.9130.602
NoneVLA Model=SmolVLA2026.010.911—
BadVLAVLA Model=OpenVLA2026.010.9060.652
State BackdoorVLA Model=SmolVLA2026.010.8930.904
TrojanRobotVLA Model=OpenVLA2026.010.8910.498
NoneVLA Model=DP2026.010.883—
State BackdoorVLA Model=DP2026.010.8750.925
TrojanRobotVLA Model=SmolVLA2026.010.8620.445
BadVLAVLA Model=SmolVLA2026.010.8590.631
NoneVLA Model=ACT2026.010.846—
TrojanRobotVLA Model=DP2026.010.8440.583
State BackdoorVLA Model=ACT2026.010.8370.942
BadVLAVLA Model=DP2026.010.8340.667
TrojanRobotVLA Model=ACT2026.010.8210.456
BadVLAVLA Model=ACT2026.010.8080.612