GameVerse dataset enables long-horizon interactive world modeling
Original title: GameVerse: A Minute-Scale Gameplay Dataset for Long-Horizon Interactive World Modeling
GameVerse contains 69.7K clips totaling 1,655 hours with stable camera trajectories for training extended interactive models.
GameVerse addresses limitations in long-horizon interactive world models by providing a dedicated dataset of minute-scale gameplay videos. It includes 69.7K clips spanning 1,655 hours and 349.6 million frames from diverse first- and third-person sources, including 1,358 hours of continuous two-minute sequences. Each clip features a unified segmentation pipeline, event-based captions separating static and dynamic elements, and coherent camera trajectories. Training on GameVerse extends the temporal coherence of interactive world models beyond one minute.
- 69.7K clips across 1,655 hours and 349.6 million frames
- 1,358 hours of continuous two-minute sequences
- Event-based caption annotation aligned to scene changes
- Surpasses prior datasets in scale and annotation quality
From the original article
1 min read at huggingface.co
Recent advances in interactive world models have made it possible to generate controllable, dynamic environments from action inputs. Yet current models struggle to maintain coherence beyond roughly one minute of rollout, a limitation rooted not in model capacity but in the scarcity of long-horizon training data.
By Kang He, Wenshuo Peng, Chuanhao Li · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.