MolmoWeb-SyntheticSkills: Web Navigation Agent Trajectories with Screenshots and Actions

MolmoWeb-SyntheticSkills is a dataset of synthetic web-navigation skills created by Allen Institute for AI (allenai). Each example pairs an instruction with a sequence of webpage screenshots and the corresponding low-level agent actions like clicks, typing, and scrolling. The dataset was last updated on March 24, 2026.

Use Cases

• Train web-navigation agents based on sequences of screenshots and actions.
• Benchmark agent performance on low-level tasks like 'find_and_open' or 'fill_form'.
• Develop multimodal models that interpret visual webpage states to predict the next action.
• Study human-computer interaction patterns from synthetic agent trajectories.

Strengths

• Pairs textual instructions with visual webpage screenshots.
• Includes sequences of agent actions like clicks, typing, and scrolling.
• Created by a known AI research institution, Allen Institute for AI (allenai).
• Last updated on March 24, 2026, suggesting recent maintenance.

Limitations

• Row count, file formats, and column-level documentation are unknown.
• Data is synthetic, which may not perfectly reflect real-world user behavior or website complexity.
• License information is not provided, which may restrict usage.

Provenance

Source

Allen Institute for AI (allenai)

Collection Method

Synthetically generated trajectories for low-level web-navigation tasks.

Time Range

null

Freshness

Last updated 2026-03-24 05:35:08

Geography

null