Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

AndroidControl

Benchmarks

Task NameDataset NameSOTA ResultTrend
GUI planningAndroidControl Low
SR (%)95.05
40
trajectory-based action predictionAndroidControl
Step Accuracy48.2
34
GUI planningAndroidControl High
SR75.8
30
GUI NavigationAndroidControl High
SR (Success Rate)76.3
29
GUI AutomationAndroidControl High
Task Match (TM)83.7
27
GUI reasoningAndroidControl Low
SR91.8
24
UI NavigationAndroidControl (offline)
Step Success Rate79.1
23
Action PredictionAndroidControl Low v2
Pass@1 Step Accuracy88.9
22
Action PredictionAndroidControl High v2
Pass@1 Step Accuracy68.59
22
GUI UnderstandingAndroidControl High
Task Match Rate (TM)83.7
22
Mobile Agent EvaluationAndroidControl Low (test)
Task Success Rate93.7
22
Step AccuracyAndroidControl High Level v2
Pass@164.3
20
Step success rateAndroidControl Task-Unseen
SSR62.3
20
GUI Interaction ControlAndroidControl (High)
SR79.37
20
General Agent CapabilityAndroidControl High
Type Rate84.7
17
High-level instruction followingAndroidControl
Step Accuracy72.7
16
Short-term planningAndroidControl Low
Type85.17
16
Action PredictionAndroidControl High-level
Click TMR87.4
15
Action PredictionAndroidControl Low-level
TMR (Click)97.6
15
General Agent CapabilityAndroidControl Low
Type Score97.2
15
Step ExecutionAndroidControl High
Step Success Rate75.3
15
Mobile Agent NavigationAndroidControl 1.0 (test)
Step Success Rate (SR)72.9
15
Mobile GUI Agent ExecutionAndroidControl Curated Easy
Type Success Rate88.6
15
Mobile GUI Agent ExecutionAndroidControl Curated-Hard
Type Rate80.8
15
Mobile UI AutomationAndroidControl High (test)
Success Rate (SR)80.3
14
Showing 25 of 60 rows