DigiData: Mobile Interaction Trajectories for Systematic App Feature Coverage

DigiData consists of mobile interaction trajectories released by Meta Research in 2025 to train mobile control agents. The data follows a specialized collection protocol designed to systematically cover application features, though the total record count is not specified in the primary metadata.

Use Cases

• Training autonomous agents to navigate mobile interfaces using episode_id to sequence state-action pairs
• Evaluating the success rate of UI automation models across the specific app features documented in the trajectories
• Fine-tuning large action models (LAMs) on mobile-specific interaction patterns

Strengths

• CC BY 4.0 open license
• Protocol-driven feature coverage
• Meta Research authorship

Limitations

• Unknown total record count
• Potential geographic bias toward US applications
• Requires external documentation for full field definitions

Provenance

Source

Meta Research (Facebook)

Collection Method

Systematic protocol-based interaction logging

Freshness

Last updated February 2026.

Geography

United States