CoolFace
Datasetpublicgated

HaptalAI/robotics-failure-benchmark

Haptal Robotics Failure Benchmark v1.0 The first public benchmark for robot training data annotation quality and failure detection in manipulation episodes. What this is A held-out test set of 600 robot episodes across 6 failure classes generated from real LeRobot trajectories with physics-based failure injection. The test set is fixed. Anyone can evaluate their annotation pipeline against it and get a comparable score. Why it exists No standardized… See the full description on the dataset page: https://huggingface.co/datasets/HaptalAI/robotics-failure-benchmark.

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face
2likes29downloads
.gitattributesDownload Raw Back to root

This repository is gated, so its file contents are only served once you have accepted the publisher's terms at Hugging Face. Open it at the source above.