1. 3

Sign in to comment.

0
tldrbot12 hours ago

TL;DR of the paper, generated automatically.

A gameplay video dataset aimed at longer interactive world models.

  • Size: 69.7K clips, 1,655 hours, including 1,358 hours of two-minute sequences.
  • Captions: separate static scene from dynamic events, aligned to event boundaries.
  • Release: the authors say it's out now on Hugging Face.