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

Android Control

Benchmarks

Task NameDataset NameSOTA ResultTrend
GUI GenerationAndroid Control (in-domain)
Sad94.28
14
GUI NavigationAndroid Control Low 12
Success Rate87.7
9
GUI NavigationAndroid Control High 12
Success Rate71.17
9
GUI Action PredictionAndroid Control High
Task Match (TM)76.5
6
GUI Action PredictionAndroid Control Low
TM94.1
6
GUI GroundingAndroid Control LowEM
Accuracy93.7
6
GUI GroundingAndroid Control (HighEM)
Accuracy67.36
6
Android GUI NavigationAndroid Control high
Success Rate50.8
5
Android GUI NavigationAndroid Control low
Success Rate67.4
5
Android GUI NavigationAndroid Control Low complexity
Success Rate87.7
5
Android GUI NavigationAndroid Control High complexity
Success Rate65.6
5
GUI ExecutionAndroid Control High Instruction Abstraction
Type Match (TM)71.9
4
GUI ExecutionAndroid Control Low Instruction Abstraction
Type Match (TM)91
4
Showing 13 of 13 rows