HuggingFaceVLA/smol-libero

Smol-LIBERO是一个LIBERO基准的紧凑版本，专为快速和易于实验设计。它包含的轨迹和摄像头数量较少，但保持了LIBERO的多模态结构。每个样本包括来自两个固定摄像头的图像、两种机器人状态（末端执行器位姿+夹爪状态和完整的7自由度关节位置）以及动作（7自由度关节命令）。这个设置特别适合比较低维状态输入和高维视觉输入，或者在多模态训练中将它们结合起来。

Smol-LIBERO is a compact version of the LIBERO benchmark, designed for fast and accessible experimentation. It contains fewer trajectories and cameras while maintaining the multimodal structure of LIBERO. Each sample includes images from two fixed cameras, two types of robot states (end-effector pose + gripper, and full 7-DoF joint positions), and actions (7-DoF joint commands). This setup is particularly useful for comparing low-dimensional state inputs with high-dimensional visual inputs, or for combining them in multimodal training.