Open, reproducible evidence

Know your robot data
before training.

Explore pinned checks of public LeRobot, MCAP, and ROS 2 data. Every result links to the exact report.

Explore the atlas ↓Submit evidence ↗
23published observations
237applicable checks
3quality profiles

Evidence atlas

Explore every check.

Filter by format, inspect the outcome, and open the reproducible report.

#Dataset / recordingProfileScaleChecksFindingsOutcomeEvidence
01Robomimic CAN PH ↗Panda · Public evidenceLeRobot10 eps · 1.2k frames110PassedReport ↗ · Link
02Scout Earth Rover Mini ↗Earth Rover Mini · Public evidenceLeRobot3 eps · 4.2k frames140PassedReport ↗ · Link
03SVLA SO-101 Pick & Place ↗SO-101 · Public evidenceLeRobot50 eps · 11.9k frames140PassedReport ↗ · Link
04Sentinel Demo 09 ↗YAM humanoid · Public evidenceLeRobot11 eps · 14.0k frames140PassedReport ↗ · Link
05Foxglove MCAP TenMessages conformance case ↗Generic channel · Public evidenceMCAP10 messages75Issues foundReport ↗ · Link
06RobotisAI ARX X5 elevator-button episode ↗ARX X5 · Public evidenceROS 27047 messages120PassedReport ↗ · Link
07Controlled ROS 2 JointState recordingROS 2 fixture · Controlled evidenceROS 23 messages120PassedReport ↗ · Link
08Controlled ROS 2 cadence and dimensions corruptionROS 2 fixture · Controlled evidenceROS 23 messages132Issues foundReport ↗ · Link
09PushT ↗2D pusher · Survey evidence · Likely frozen-frame false positive on a low-texture 2D sim.LeRobot206 eps · 25.6k frames1550Issues foundReport ↗ · Link
10xArm Lift Medium ↗xArm · Survey evidenceLeRobot800 eps · 20.0k frames150PassedReport ↗ · Link
11Unitree H1 Two-Robot Greeting ↗Unitree H1 · Survey evidenceLeRobot30 eps · 3.8k frames150PassedReport ↗ · Link
12SVLA SO-100 Pick & Place ↗SO-100 · Survey evidenceLeRobot50 eps · 19.6k frames150PassedReport ↗ · Link
13Aloha Static Battery (episodes 0-4) ↗ALOHA · Survey evidenceLeRobot5 eps · 3.0k frames150PassedReport ↗ · Link
14UR10e Robotiq 2F-85 ↗UR10e + Robotiq · Survey evidence · Four 1-second holds during motion; review, not a gate claim.LeRobot55 eps · 21.4k frames154Issues foundReport ↗ · Link
15SO-101 Teleop Pick-and-Place ↗SO-101 · Survey evidence · Wrist-camera freezes at the 50-finding cap; likely static close-ups.LeRobot50 eps · 18.2k frames1550Issues foundReport ↗ · Link
16Reachy 2 Head Sample ↗Reachy 2 · Survey evidence · Confirmed defect: episode metadata names parquet shards the Hub does not publish.LeRobot10 eps · 3.1k frames53Issues foundReport ↗ · Link
17LeKiwi Go2 ↗LeKiwi · Survey evidenceLeRobot40 eps · 14.4k frames150PassedReport ↗ · Link
18LIBERO Spatial ↗Franka · Survey evidence · One 6-frame freeze at episode start; review, not a gate claim.LeRobot432 eps · 53.0k frames151Issues foundReport ↗ · Link
19Koch arm gripper (v2.1) ↗Koch · Survey evidence · Koch embodiment is published as v2.1; keep as fail-closed evidence.LeRobotv2.1 metadata00RejectedReport ↗ · Link
20Koch pick-and-place Lego (v2.0) ↗Koch · Survey evidence · Official-style Koch teleop remains v2.0 on the Hub.LeRobotv2.0 metadata00RejectedReport ↗ · Link
21SO-101 pi0.5 base (v2.1) ↗SO-101 · Survey evidence · Large community SO-101 corpus still on v2.1; metadata-only screen.LeRobotv2.1 metadata00RejectedReport ↗ · Link
22Reachy 2 kitchen multimodal (v2.0) ↗Reachy 2 · Survey evidence · Full Reachy kitchen set is v2.0; the v3 head sample is in the check list.LeRobotv2.0 metadata00RejectedReport ↗ · Link
23RoboTwin 3D (v2.1) ↗ALOHA / RoboTwin · Survey evidence · Bimanual Aloha conversion at 50 FPS remains v2.1.LeRobotv2.1 metadata00RejectedReport ↗ · Link

Order is the catalog, not a ranking. Release-gate rows are the public claim. Survey rows are compatibility observations and may include classified false positives.

Selected findings

Concrete defects. Exact evidence.

Each finding links to its rule and sanitized report.

01Confirmed public defect

Metadata pointed to shards that were never published.

The Reachy 2 Head sample declared Parquet files that were absent from its public revision. The dataset could be discovered, but not reliably consumed.

manifest.required_filesRead evidence ↗
02Controlled evidence

One reordered timestamp exposed three timing failures.

A deterministic edit reproduced clock reversal, an unexpected sampling interval, and an excessive gap, without changing the original source.

temporal.monotonicRead evidence ↗
03ROS 2 evidence

A topic gap and malformed JointState were separated cleanly.

The recording showed both a cadence break and a position/name dimension mismatch, each reported under the rule that owns it.

ros2.semantic_consistencyRead evidence ↗

Contribute

Add your robot data.

Share a pinned public dataset or recording. Physlint keeps the source untouched and publishes only sanitized evidence.

Submit evidence ↗Discuss ↗
  1. 01

    Share a revision

    Link a public source that can be pinned.

  2. 02

    Add context

    Name the robot, format, and expected behavior.

  3. 03

    Get evidence

    Receive a scoped, reproducible report.

Run it locally

Check data before training.

pip install physlintRun Physlint ↗