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

AndroidWorld

Benchmarks

Task NameDataset NameSOTA ResultTrend
GUI Agent TaskAndroidWorld
Success Rate80
200
Mobile Task AutomationAndroidWorld (test)
Average Success Rate1
119
GUI AgentAndroidWorld
Accuracy62
70
Mobile GUI AutomationAndroidWorld
Overall Success Rate76.9
68
GUI navigationAndroidWorld latest (test)
Success Rate78.9
48
Mobile UI ControlAndroidWorld
Overall Task Success Rate71.6
45
Mobile GUI Agent Decision MakingAndroidWorld
Success Rate82.05
27
GUI Agent Task SuccessAndroidWorld (online)
Task Success Rate48.7
25
End-to-end GUI NavigationAndroidWorld
Success Rate77.6
25
Android GUI AutomationAndroidWorld M
L1 Score53.8
22
Agentic ReasoningAndroidWorld
Success Rate82.05
20
Mobile GUI AgentsAndroidWorld 138 tasks (test)
Success Rate71.1
18
Mobile UI Task AutomationAndroidWorld (full)
Success Rate (Easy)82.79
16
Mobile Agent Decision-makingAndroidWorld (Evaluation set 116 templates)
Average Success Rate (SR)62.9
16
GUI Task CompletionAndroidWorld
Success Rate (SR)57.8
15
Reward ModelingAndroidWorld
Precision92.5
14
End-to-End Environment InteractionAndroidWorld (test)
Pass@180.2
14
Mobile agent task completionAndroidWorld 116 tasks
Task Success Rate (TSR)75
10
Online Reinforcement LearningAndroidWorld
Success Rate62.28
9
Agentic Phone UseAndroidWorld
Task Success Rate83.2
7
GUI Agent AutomationAndroidWorld (AW) (Online)
Success Rate25.1
6
Safe NavigationAndroidWorld core20 safe general tasks
Success Count (out of 20)11
4
Mobile UseAndroidWorld
Score70.7
4
Mobile operating system task executionAndroidWorld (AW)
AUV43.2
4
Action Success PredictionAndroidWorld (smallest held-out split)
Precision87.1
3
Showing 25 of 27 rows