AndroidControl
A public benchmark for GUI agents on Android devices. It contains screenshots from real applications and human operation trajectories, evaluating an agent's ability to predict the next action from a natural language instruction.
Timeline (last 90 days)
On the AndroidControl benchmark, GUI-JEPA's inference speed is 2.8 to 4.0 times that of a Qwen2.5-VL-7B agent fine-tuned with the same input.
Using only about 6% of Qwen2.5-VL-7B's parameter count, GUI-JEPA reached 86% to 92% of the latter's in-distribution accuracy on AndroidControl.
AndroidControl is a public benchmark for GUI agents on Android devices, containing real app screenshots and human operation trajectories, with tasks given as natural language instructions and evaluation typically distinguishing in-distribution and out-of-distribution settings.
GUI-JEPA has so far been validated only on AndroidControl, its action space and task types are relatively limited, and its cross-platform generalization ability has not yet been fully verified.
All Facts (4)
On the AndroidControl benchmark, GUI-JEPA's inference speed is 2.8 to 4.0 times that of a Qwen2.5-VL-7B agent fine-tuned with the same input.
50%UnverifiedUsing only about 6% of Qwen2.5-VL-7B's parameter count, GUI-JEPA reached 86% to 92% of the latter's in-distribution accuracy on AndroidControl.
50%UnverifiedAndroidControl is a public benchmark for GUI agents on Android devices, containing real app screenshots and human operation trajectories, with tasks given as natural language instructions and evaluation typically distinguishing in-distribution and out-of-distribution settings.
50%UnverifiedGUI-JEPA has so far been validated only on AndroidControl, its action space and task types are relatively limited, and its cross-platform generalization ability has not yet been fully verified.
50%