Android in the Wild: Human Demonstrations for Device Control

Android in the Wild (AITW) is a large-scale dataset for Android device control containing human demonstrations. It is a mirror of Google's original dataset, re-hosted on Hugging Face by user leosltl for easier community access. The dataset page was last updated on April 23, 2026.

Use Cases

• Training reinforcement learning agents for mobile navigation based on human demonstration trajectories.
• Developing models for on-screen action prediction based on recorded human interactions.
• Benchmarking autonomous Android device control agents against a large-scale corpus of real-world tasks.

Strengths

• Dataset is described as 'large-scale', suggesting a substantial number of demonstrations.
• Directly mirrors a dataset from Google Research, indicating a credible original source.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Google Research (original), mirrored by leosltl on Hugging Face.

Collection Method

Contains human demonstrations of Android device control.

Freshness

Last updated 2026-04-23 03:09:58; freshness should be verified.