Benchmark tests robots on quick safety reactions to household hazards
ReactHuman: A Physics-Grounded Benchmark for Human-Like Reactive Decision-Making in Embodied Multimodal LLMs
RoboticsArtificial Intelligence
Summary
Reacting quickly to unexpected dangers like slipping or falling objects is a hard problem for robots that need to act safely at home. The authors created ReactHuman, a test where simulated robots face sudden household hazards and must react correctly based on real physical rules. They tested several large multimodal language models and found that current models often fail to respond safely or logically, sometimes relying on wrong clues like appearance over actual motion. This benchmark helps researchers understand these failures and improve robots’ immediate reactive decision-making.
What this means in practice
- •For household robot developers: Evaluate and improve reactive safety behaviors of robots facing sudden physical hazards in home environments using a physics-grounded test suite.
- •For robotics safety teams: Use precise simulation data to diagnose and reduce unsafe or unreasonable actions in robots interacting with dynamic household objects.
Authors
Yizhan Li, Jianxin You, Mengyang Xiong, Yinhuan Chen, Zicheng Zhao, Dekun Wu, Dongqing Zhang, Bang Liu
Abstract
Reacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision coreof household robots. Existing evaluations, however, probe intuitive physics passively through question answering over videos, or target deliberate, long-horizon tasks such as navigation and rearrangement; none measure whether a model can turn physical understanding into immediate, safety-critical action. We introduce ReactHuman, the first physics-grounded benchmark for human-like reactive decision-making, in which the evaluated MLLM acts as the brain of a simulated humanoid facing sudden household hazards; it spans 17 event families and over 1,000 bit-for-bit reproducible scenes with exact, annotation-free ground truth derived from 240 Hz rigid-body simulation, including adversarial objects whose appearance contradicts their physics (a foam anvil, a steel apple). We further design a five-metric suite that scores each reaction along three axes: reasonable, safe, and physically grounded. We physically execute every committed plan so that decisions have observable consequences. With this harness we evaluate seven representative MLLMs. Results show that reactive safety is far from solved: models mishandle roughly one hazard in three, act from fixed dispositions rather than the observed scene, trust appearance over motion, and miss interception points at meter scale even when the chosen action is correct; none of these failures shrink with model scale. ReactHuman thus offers both a fine-grained diagnosis and a scalable training signal toward physically grounded, safety-aware embodied agents. The benchmark can be found here: https://huggingface.co/datasets/Alan123/reacthuman-benchmark-scaled