ROBOSHACKLES: A Safety Dataset for Human-Injury Prevention in Embodied Foundation Models
Zhuowen Yin, Chongyang Liu, Wenzhang Yang, Renjue Li, Yinxing Xue
THE PROBLEM
This paper focuses on learning. This paper reveals that current embodied foundation models (like those powering Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Control & PlanningControlThe method used to make the robot move the way you want. systems) generate unsafe actions 100% of the time in hazardous scenarios—and provides a 10,000-clip Robot LearningDatasetA collection of training or evaluation data. to train them to refuse dangerous tasks. If you're building Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems with foundation models, this shows a critical gap and gives you Simulation & Sim-to-RealBenchmarkA standard test used to compare methods fairly. data to measure and improve safety guardrails. Read the paper by tracking the Core ConceptsTaskThe job the robot is supposed to complete, such as pick-and-place, navigation, or drawer opening. definition, the Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. or data assumptions, and the evidence that supports the claimed improvement.
HOW IT WORKS
Task framing
Core method
Data and supervision
Evaluation evidence
KEY RESULTS
This paper reveals that current embodied foundation models (like those powering Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Control & PlanningControlThe method used to make the robot move the way you want. systems) generate unsafe actions 100% of the time in hazardous scenarios—and provides a 10,000-clip Robot LearningDatasetA collection of training or evaluation data. to train them to refuse dangerous tasks. If you're building Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems with foundation models, this shows a critical gap and gives you Simulation & Sim-to-RealBenchmarkA standard test used to compare methods fairly. data to measure and improve safety guardrails.
WHY DEVELOPERS SHOULD CARE
This paper reveals that current embodied foundation models (like those powering Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. Control & PlanningControlThe method used to make the robot move the way you want. systems) generate unsafe actions 100% of the time in hazardous scenarios—and provides a 10,000-clip Robot LearningDatasetA collection of training or evaluation data. to train them to refuse dangerous tasks. If you're building Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. systems with foundation models, this shows a critical gap and gives you Simulation & Sim-to-RealBenchmarkA standard test used to compare methods fairly. data to measure and improve safety guardrails.
LIMITATIONS
The main limitation to check is whether the claimed behavior holds outside the paper's reported setup. That means testing across different Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. embodiments, scenes, objects, and data distributions.
WHAT COMES NEXT
The practical next step is independent reproduction with clear baselines, ablations, and stress tests. For a developer, the useful follow-up is to map the paper's learning assumptions onto a concrete Core ConceptsRobotA physical system with sensors and actuators that can observe the world and take actions. stack, then test the smallest version of the method that could run end to end.