UI-MOPD collected 2,640 agent interaction trajectories comprising approximately 25,900 steps on Android mobile devices. The data was gathered via AndroidWorld from 19 open-source Android applications and used for training the UI-MOPD model. The dataset was last updated on July 5, 2026.
Use Cases
- Training mobile GUI automation agents based on recorded interaction trajectories.
- Benchmarking on-policy distillation methods for continual learning based on the described multi-platform training data.
- Studying cross-application GUI interaction patterns based on data from 19 different open-source apps.
Strengths
- Contains 2,640 distinct agent interaction trajectories.
- Includes approximately 25,900 total interaction steps.
- Covers 19 different open-source Android applications.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Freshness should be verified as the last update was July 5, 2026.
Provenance
- Source
- UI-MOPD
- Collection Method
- Collected on open-source Android applications via AndroidWorld.
- Freshness
- Last updated 2026-07-05 03:52:50