zrgong/rmbench-plus-plus-press-button-update

RMBench++ Press Button Update v1.1

This dataset contains 100 RT-rendered scripted-expert demonstrations for the RMBench++ press_button_update training split.

Each number card gives the required registered-press total for the counting button below it. The left arm operates both counting buttons and the right arm confirms once. During execution, one visible target may increase, so the policy must keep observing the cards, continue from presses already registered, and confirm only when both current targets are satisfied. The first registered confirm edge is irreversible.

Canonical task prompt

Each number card shows how many successful presses its counting button requires. Use the left arm for both counting buttons. Watch the number cards and indicator lights throughout the task, and count only presses acknowledged by the corresponding light. When both successful-press totals match the current targets, confirm once with the right arm.

The same prompt is used verbatim for training and evaluation; only the frozen seed ranges differ. Version v1.1.0 replaces the prompt paraphrases embedded in v1.0.0. Robot trajectories and videos are unchanged.

Dataset contents

• 100 episodes using the frozen training seeds 1 through 100.
• 62,942 frames at 30 FPS.
• One Parquet file per episode and one H.264 MP4 per episode per camera.
• Native RMBench 14D observation.state and 14D action: six left-arm joints, left gripper, six right-arm joints, right gripper.
• Policy views: head_camera, left_camera, and right_camera, each RGB 320 x 240. front_camera is included for QA and visualization and is not a default policy input in the v1.1 protocol.
• Per-frame episode seed, public prompt provenance, and JSON-encoded rmbench_plus audit metadata.

The embedded meta/rmbench_plus_manifest.json records the protocol version and the disjoint training/evaluation seed ranges. This repository contains training demonstrations only; it does not contain held-out evaluation rollouts.

Loading

from lerobot.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset(
    repo_id="zrgong/rmbench-plus-plus-press-button-update",
    video_backend="pyav",
)

Validation and provenance

The uploaded tree passed the RMBench++ LeRobot v3 validator with:

• codebase_version == "v3.0"
• 100 episode metadata rows
• 100 data Parquet files
• 400 video files across four camera keys
• 14D state and action reads through LeRobotDataset
• exact seeds 1 through 100
• 100/100 scripted-expert update events and confirmations successful

The canonical prompt contract, manifest, converter, and validator are pinned at RMBench++ code commit b36949dcb0d2cf0ca53427c9f28d2aec0f4c97ab.

These demonstrations are qualitative and training evidence from the scripted expert. They are not closed-loop policy evaluation results. RMBench++ v1.1 is a pilot protocol, not a general-memory leaderboard.