Action-conditioned safety-judgment data from ManiGuard rollouts across six task families. Inputs: adjacent overview/wrist images, current robot state, the masked unexecuted action suffix, and a natural-language constraint. Targets indicate a new constraint violation during that suffix; monitor outputs are supervision only. The delivery contains raw/ episode records and prepared package/ indices. Training caches are built on the training machine. See SafetyJev training instructions. Preparation in progress. Use the release only when DATASETSUMMARY.json reports complete. Access for external users requires manual approval.
Organization
IDEAS-Lab-Northwestern
IDEAS-Lab-Northwestern
Models in Library0
Datasets in Library1
Models on Hugging Face87
Followers6
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.