unknown

GameplayQA

A dense benchmark for first-person multi-video understanding of 3D agents, featuring 1.22 labels/second annotations and 2.4K diagnostic QA pairs across a Self-Other-World taxonomy.

<p>GameplayQA

A dense benchmark for first-person multi-video understanding of 3D agents, featuring 1.22 labels/second annotations and 2.4K diagnostic QA pairs across a Self-Other-World taxonomy.</p>

GameplayQA is a dense benchmark designed for first-person multi-video understanding of 3D agents, featuring a high annotation density of 1.22 labels per second. It targets evaluation of AI models on gameplay video comprehension tasks.