Dataset · MLL-Lab
Simulation robotics benchmark and agent logs
MLL-Lab/embodied-exam-data
Media data and code agent run logs for the Embodied First Exam robot simulation benchmark.
What's inside
The dataset contains between 100 thousand and 1 million elements: images, video recordings, and complete agent attempt logs from the Embodied First Exam leaderboard. Each task represents a simulator scene with a text goal, and the answer is program code verified by rerunning in a clean simulation environment.
What you can do with it
- Evaluating and comparing code agents for embodied AI and robotics tasks
- Analyzing trajectories and errors of AI agents based on published leaderboard runs
- Size
- 1K–10K rows
- Downloads this month
- 41.3K
- Likes
- 0
How to load it
from datasets import load_dataset
ds = load_dataset("MLL-Lab/embodied-exam-data")