DexJoCo: A Benchmark for Task-Oriented Dexterous Manipulation

DexJoCo provides a benchmark and toolkit dataset for task-oriented dexterous manipulation research. The dataset is formatted for the LeRobot framework and is associated with a 2026 research paper. It was created by the DexJoCo organization and last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking reinforcement learning agents on dexterous manipulation tasks based on the described benchmark.
• Training robotic control policies for tasks requiring fine motor skills based on the task-oriented focus.
• Developing and evaluating new algorithms within the LeRobot framework based on the provided format.

Strengths

• Dataset is explicitly designed as a benchmark, suggesting a standardized evaluation framework.
• Associated with a detailed research paper (arXiv:2605.16257) providing context and methodology.
• Formatted for the LeRobot framework, which may ease integration with related tools.

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 size are unknown, which may limit suitability assessment.

Provenance

Source

DexJoCo organization, associated with research paper.

Collection Method

Likely generated via simulation in the MuJoCo physics engine.

Freshness

Last updated 2026-05-20 07:01:19; freshness should be verified.