WEB-Dataset: 90 Everyday Bimanual Manipulation Tasks with Language Annotations

WorldEngineAI's WEB-Dataset is a large-scale, language-annotated real-robot bimanual manipulation dataset intended for post-training robotics foundation models. It spans 90 everyday manipulation tasks collected with a bimanual YAM follower arm teleoperated by a GELLO leader. The dataset records joint state, action, and three synchronized camera streams at 60 Hz.

Use Cases

• Post-training or fine-tuning robotics foundation models based on real-robot bimanual manipulation data.
• Developing imitation learning policies based on teleoperated joint state and action sequences.
• Training vision-language-action models based on synchronized camera streams and language annotations.
• Benchmarking bimanual manipulation performance across 90 diverse everyday tasks.

Strengths

• Covers 90 distinct everyday manipulation tasks, suggesting breadth.
• Records data at 60 Hz, which likely provides high temporal resolution.
• Includes three synchronized camera streams, offering multiple visual perspectives.
• Contains language annotations, which may facilitate instruction-following research.

Limitations

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

Provenance

Source

WorldEngineAI

Collection Method

Teleoperated collection using a bimanual YAM follower arm controlled by a GELLO leader.

Time Range

null

Freshness

Last updated 2026-07-12 14:30:26; freshness should be verified.

Geography

null