turing-motors/ACT-Bench

ACT-Bench是一个用于定量评估自动驾驶中世界模型动作可控性的专用框架。它专注于测量世界模型在指定轨迹条件下生成驾驶场景的能力。数据集包含图像和轨迹数据，用于生成指定轨迹的驾驶场景，并评估模型的性能。

ACT-Bench is a dedicated framework for quantitatively evaluating the action controllability of world models for autonomous driving. It focuses on measuring how well a world model can generate driving scenes conditioned on specified trajectories. The dataset includes image and trajectory data used to generate driving scenes conditioned on specified trajectories and to evaluate model performance.