List your company

Mobile GUI Interaction on Android GUI Evaluation Benchmark 500 human-annotated trajectories (test)

78.55AccuracyMobile-R1
53.527660.023866.5273.0162Jun 25, 2025

Evaluation results

MethodLinks
Mobile-R1Training Stages=Stage 1, 2, & 32025.0678.5530.637.4241
Mobile-R1Training Stages=Stage 1 & 22025.0677.6929.436255
Qwen2.5-VL-32BBackbone=Qwen2.5-VL, Parameters=32B2025.0675.930.4—280
Mobile-R1Training Stages=Stage 12025.0675.6824.429.8280
GUI-R1-7BParameters=7B2025.0671.7232.6—298
AgentCPM-8BParameters=8B2025.0671.6530—338
Qwen2.5-VL-7BBackbone=Qwen2.5-VL, Parameters=7B2025.0663.4612.821.6523
GUI-R1-3BParameters=3B2025.0661.2912—461
UI-R1-3B-EParameters=3B2025.0659.129.4—473
UI-R1-3BParameters=3B2025.0656.1317.2—451
Qwen2.5-VL-3BBackbone=Qwen2.5-VL, Parameters=3B2025.0654.497.216.33651