Miical/libero10-agent-500demos

This dataset was created using LeRobot.

Dataset Description

LIBERO-10 Agent 500 is a collection of 500 successful robot-manipulation demonstrations covering all 10 LIBERO-10 tasks. Each task contains one complete successful trajectory for each of its 50 official initial states. The dataset is exported in native LeRobot v3 format for imitation learning, supervised fine-tuning, and robotics post-training experiments.

At a glance

• 500 episodes, balanced across 10 tasks (50 episodes per task).
• 174,873 action frames recorded at 20 FPS.
• Panda robot with two 256 × 256 RGB camera streams: scene and wrist.
• 8-dimensional observations (observation.state) and 7-dimensional actions.
• Natural-language task instructions, rewards, termination flags, success flags, and original reset-state IDs are included.

Task coverage

Task ID	Task	Episodes
0	Put the alphabet soup and tomato sauce in the basket	50
1	Put the cream cheese and butter in the basket	50
2	Turn on the stove and put the moka pot on it	50
3	Put the black bowl in the bottom drawer and close it	50
4	Put the white mug on the left plate and the yellow-and-white mug on the right plate	50
5	Place the book in the back compartment of the caddy	50
6	Put the white mug on the plate and the chocolate pudding to its right	50
7	Put the alphabet soup and cream cheese in the basket	50
8	Put both moka pots on the stove	50
9	Put the yellow-and-white mug in the microwave and close it	50

How the demonstrations were collected

The demonstrations were collected through the verl-vla agent-collection workflow, using a scripted external planner with privileged simulator-state feedback. The planner issued robot actions, inspected observations, and could restore recorded checkpoints to revise unsuccessful action sequences. Some trajectories include clearing an obstructing object before completing the task.

Only simulator-confirmed successful branches are included. Each exported episode retains the complete trajectory from its initial state, including the prefix before any checkpoint restoration. Failed branches and diagnostic comparison videos are retained in the original collection archive, but are not part of this Hugging Face dataset. Task 8 reuses 50 demonstrations from the preceding collection run; the remaining 450 were collected for this release.

These are privileged-state-assisted demonstrations, not human teleoperation recordings or evaluations of a learned policy. The collection's successful terminal states do not establish a 100% policy success rate or held-out generalization. All 500 episodes are in the training split; no separate validation or test split is provided.

Episode mapping and validation

info.reset_state_id = task_id * 50 + initial_state_index, with zero-based task IDs and initial-state indices. Episodes are sorted by this ID, so episode_index equals info.reset_state_id in this release.

Before publication, all episodes were checked for finite states/actions, continuous frame indices, correct timestamps, and successful terminal transitions. Both camera videos for every episode were fully decoded and checked against the action-frame count. Merged data values and camera files were verified against the successful source exports, and all 1,000 episode endpoint samples were read successfully through the native LeRobot API.

Loading the dataset

from lerobot.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset(
    "Miical/libero10-agent-500demos",
    video_backend="pyav",
)
sample = dataset[0]

The camera keys are observation.images.image and observation.images.wrist_image. Use the Visualize this dataset button above to inspect episodes interactively.

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "robot_type": "panda",
    "total_episodes": 500,
    "total_frames": 174873,
    "total_tasks": 10,
    "chunks_size": 1000,
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 1e-06,
    "fps": 20,
    "splits": {
        "train": "0:500"
    },
    "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
    "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
    "features": {
        "observation.images.image": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "channel"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "video.fps": 20,
                "video.channels": 3,
                "has_audio": false
            }
        },
        "observation.images.wrist_image": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "channel"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "video.fps": 20,
                "video.channels": 3,
                "has_audio": false
            }
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                8
            ],
            "names": [
                "state"
            ]
        },
        "action": {
            "dtype": "float32",
            "shape": [
                7
            ],
            "names": [
                "action"
            ]
        },
        "next.reward": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null
        },
        "next.terminated": {
            "dtype": "bool",
            "shape": [
                1
            ],
            "names": null
        },
        "next.truncated": {
            "dtype": "bool",
            "shape": [
                1
            ],
            "names": null
        },
        "next.success": {
            "dtype": "bool",
            "shape": [
                1
            ],
            "names": null
        },
        "info.is_intervention": {
            "dtype": "bool",
            "shape": [
                1
            ],
            "names": null
        },
        "info.reset_state_id": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "timestamp": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null
        },
        "frame_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "episode_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "task_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        }
    }
}

Citation

BibTeX:

[More Information Needed]