Skip to content

Dataset · IDEAS-Lab-Northwestern

Home robotics safety benchmark

IDEAS-Lab-Northwestern/ManiGuard-Bench

1,000 simulations to verify how accurately and safely robots perform household tasks.

What's inside

Contains 1,000 test scenarios across 6 families of household tasks (200 basic tasks) in simulation. Each scenario includes a saved scene, text instruction, goal description, safety time constraints, as well as images and video (five 640x640 camera angles and reference frames). The dataset is designed for separate evaluation of action success and adherence to safety rules.

What you can do with it

  • Testing multimodal models (VLA) for household robot control
  • Evaluating compliance with safety constraints during action planning in simulation
Tasks
Robotics
Size
1K–10K rows
Downloads this month
45.8K
Likes
2

How to load it

from datasets import load_dataset

ds = load_dataset("IDEAS-Lab-Northwestern/ManiGuard-Bench")