The Bleeding Edge Gameplay Sample dataset was released by Microsoft Research in 2025. It includes operation trajectories, video frames, and corresponding action data for the multiplayer online game "Bleeding Edge," aiming to provide high-quality synthetic gaming data for reinforcement learning and computer vision research.

This dataset comprises 1,024 sixty-second gameplay clips (total size approximately 75 GB), preprocessed into videos sampled at 10 fps with a resolution of 300x180 pixels.

Dataset Composition

The dataset primarily consists of test set data used for evaluation, structured as follows:

Data Files

• testing_dataset_part1.zip & testing_dataset_part2.zip: Contain all 1,024 sixty-second gameplay trajectory datasets intended for model evaluation.

Data Instances

Each data instance typically consists of one .npz file and one .mp4 video file:

• .npz File: Contains decompressed video frames, human player actions (actions), and game states (states). Specifically, the dimensions are (610, 16) for actions, (610, 56) for states, and (610, 3, 180, 300) for images.
• .mp4 File: A 60-second video that corresponds to the contents of the .npz file, facilitating intuitive visual inspection.