Issue 01AI 리서치
AC POST
AI 리서치 목록
arxiv2026년 10월 31일 09:00

GameVerse: A Minute-Scale Gameplay Dataset for Long-Horizon Interactive World Modeling

기존 인터랙티브 월드 모델이 1분 이상의 롱 호라이즌 생성에서 일관성을 유지하지 못하는 문제를 데이터 부족에서 기인한 것으로 진단했습니다. 이를 해결하기 위해 69.7K개의 클립과 1,655시간의 분 단위 게임플레이 영상을 포함하는 GameVerse 데이터셋을 구축했습니다. 이 데이터셋은 연속적인 2분 시퀀스를 포함하여 기존 데이터셋보다 한 자릿수 이상 긴 롱 호라이즌 학습 데이터를 제공합니다.

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. Existing gameplay video datasets typically consist of clips under ten seconds, with captions that describe whole clips in coarse and undifferentiated text, which is insufficient for controllable long-horizon generation. We address this with GameVerse, a minute-scale gameplay video dataset purpose-built for long-horizon interactive world modeling. GameVerse contains 69.7K clips spanning 1,655 hours and 349.6 million frames across diverse first- and third-person gameplay, including 1,358 hours of continuous two-minute sequences, an order of magnitude longer than prior datasets. Clips pass through a unified pipeline of scene-consistent segmentation, quality filtering, camera pose