MolmoAct2-LIBERO: Merged Robot Learning Episodes for Finetuning

A merged dataset of 1,693 robot episodes and 273,465 frame-level training examples, created by AllenAI using the LeRobot framework. It combines four LIBERO suites for multitask robot learning. The dataset was last updated on May 10, 2026.

Use Cases

• Finetuning robot manipulation models based on the combined LIBERO suites.
• Training multitask robotic agents based on the 273,465 frame-level examples.
• Benchmarking imitation learning algorithms based on the episode structure.
• Developing policies for long-horizon tasks based on the LIBERO-Long suite.

Strengths

• Contains 1,693 distinct robot learning episodes.
• Provides 273,465 frame-level training examples at 10 FPS.
• Combines four established LIBERO suites (Spatial, Goal, Object, Long).

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.
• Intended for finetuning only, as stated in the description.

Provenance

Source

AllenAI, created using the LeRobot framework.

Collection Method

Merged from four LIBERO suites (LIBERO-Spatial, LIBERO-Goal, LIBERO-Object, LIBERO-Long).

Time Range

null

Freshness

Last updated 2026-05-10 22:28:33; freshness should be verified.

Geography

null