# GuardianFail-36k and Guardian FailCoT OOD sets

From the catalogue. Listed by one researcher from one source page.

- id: `guardianfail-36k-and-guardian-failcot-ood-sets`
- kind: language and reasoning annotation
- robot or device: RLBench simulator and BridgeDataV2 real robot (class: cross embodiment)
- how collected: generated planning and execution failures
- size as stated: RLBench-Fail 12,358/1,000/1,000 and 5,808/500/500 splits; RoboVQA OOD 357 questions
- hours: not stated | hours only claimed: none | episodes: not stated
- year: 2025
- organisation: paulpacaud (HF account)
- licence: apache-2.0 (class: open)
- access: open (class: open)
- link (dataset): https://huggingface.co/datasets/paulpacaud/Guardian-FailCoT-OOD-datasets
- Hugging Face repository: `paulpacaud/Guardian-FailCoT-OOD-datasets`, 138 downloads in 30 days, https://huggingface.co/datasets/paulpacaud/Guardian-FailCoT-OOD-datasets
- note: Multi-view failure reasoning with chain of thought.

Confirm the licence at the link before relying on it.