Datasets:
Dataset Description
StaticEmbodiedBench is a dataset for evaluating vision-language models on embodied intelligence tasks, as featured in the OpenCompass leaderboard
(https://staging.opencompass.org.cn/embodied-intelligence/rank/brain).
It covers three key capabilities:
- Macro Planning: Decomposing a complex task into a sequence of simpler subtasks.
- Micro Perception: Performing concrete simple tasks such as spatial understanding and fine-grained perception.
- Stage-wise Reasoning: Deciding the next action based on the agent’s current state and perceptual inputs.
Each sample is also labeled with a visual perspective:
- First-Person View: The visual sensor is integrated with the agent, e.g., mounted on the end-effector.
- Third-Person View: The visual sensor is separate from the agent, e.g., top-down or observer view.
This release includes 200 open-source samples from the full dataset, provided for public research and benchmarking purposes.