List your company

GUI Element Injection Attack Success on ScreenSpot Pro (885-sample split)

100Attack Success Rate (ASR)GO-opt.
84.368888.426992.48596.5431Apr 9, 2026

Evaluation results

MethodLinks
GO-opt.Victim=GUI-Owl-7B, pass@k=32026.0410024.1835.5944.2948.3650.96
Rand. Inject.Victim=GUI-Owl-7B, pass@k=32026.0498.198.5213.5817.0320.0223.36
UT-opt.Victim=UI-Venus-1.5-8B, pass@k=32026.0496.3813.4820.2825.0928.2530.95
GO-opt.Victim=UI-Venus-1.5-8B, pass@k=32026.0496.3813.1320.8725.4429.4332.94
Rand. Inject.Victim=UI-Venus-1.5-8B, pass@k=32026.0496.042.944.716.3588.94
UT-opt.Victim=GUI-Owl-7B, pass@k=32026.0495.5923.8834.8742.6747.6451.65
GO-opt.Victim=EvoCUA-32B, pass@k=32026.0494.4614.4724.6430.6235.4138.64
Rand. Inject.Victim=EvoCUA-32B, pass@k=32026.0494.125.288.2810.4411.6413.81
UT-opt.Victim=EvoCUA-32B, pass@k=32026.0494.1214.2924.3730.9735.0537.58
Rand. Inject.Victim=GUI-Owl-32B, pass@k=32026.0493.455.28.8310.412.5814.15
Rand. Inject.Victim=EvoCUA-8B, pass@k=32026.0492.546.4710.2612.715.0217.34
UT-opt.Victim=EvoCUA-8B, pass@k=32026.0492.217.1625.8632.637.0139.95
GO-opt.Victim=EvoCUA-8B, pass@k=32026.0492.218.1427.4532.9737.6240.07
UT-opt.Victim=GUI-Owl-32B, pass@k=32026.0491.9818.9229.4839.0744.3548.65
GO-opt.Victim=GUI-Owl-32B, pass@k=32026.0491.316.5829.3336.2643.6948.02
Rand. Inject.Victim=OpenCUA-7B, pass@k=32026.0489.948.7913.9417.0920.2323.37
UT-opt.Victim=OpenCUA-7B, pass@k=32026.0489.623.3333.6741.9947.5451.58
GO-opt.Victim=OpenCUA-7B, pass@k=32026.0489.623.0834.840.8648.0551.83
GO-opt.Victim=OpenCUA-32B, pass@k=32026.0488.1425.7740.5148.7254.3658.97
Rand. Inject.Victim=OpenCUA-32B, pass@k=32026.0487.469.316.0221.0625.8429.33
UT-opt.Victim=OpenCUA-32B, pass@k=32026.0487.4624.5536.8245.9952.0756.46
UT-opt.Victim=UI-Venus-1.5-2B, pass@k=32026.0485.4215.8724.4730.1633.3335.58
Rand. Inject.Victim=UI-Venus-1.5-2B, pass@k=32026.0485.087.0410.3613.5515.0117
GO-opt.Victim=UI-Venus-1.5-2B, pass@k=32026.0484.9716.2224.630.0534.0435.9