SATraj-OS: Large-Scale GUI Trajectory Data for Computer-Using Agents

SATraj-OS is a large-scale dataset of graphical user interface (GUI) interaction trajectories for Computer-Using Agents (CUAs). The dataset is designed for both capability learning and safety training of agents that perform tasks like browsing websites, editing documents, and managing files. It was created by AI45Research and last updated on Hugging Face in April 2026.

Use Cases

• Training agents for task execution based on GUI interaction trajectories mentioned in the description.
• Evaluating agent safety and robustness using the safety training data described.
• Developing models for cross-application agent generalization based on the multi-task trajectory data.

Strengths

• Dataset is described as 'large-scale', suggesting a substantial collection of trajectories.
• Explicitly designed for dual purposes: capability learning and safety training for autonomous agents.
• Last update timestamp (2026-04-22 09:37:27) is provided, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• The full description is truncated, requiring a visit to the dataset page for complete details.

Provenance

Source

AI45Research via Hugging Face.

Collection Method

Likely contains recorded sequences of GUI interactions (trajectories) for agent training.

Freshness

Last updated 2026-04-22 09:37:27.