内容库 / 论文

arXiv:2608.24680 · 工程化

Game2World Engine: Unlocking In-the-Wild Gameplay Videos for World Model Training

Game2World Engine: Unlocking In-the-Wild Gameplay Videos for World Model Training

类型：arxiv
标识：2608.24680
链接：https://arxiv.org/abs/2608.24680
主分类：engineering
形态：method
TLDR：Video games provide a scalable source of training data for video world models, offering diverse environments, complex interactions, and abundant in-the-wild gameplay videos. However, raw gameplay footage entangles the game world with screen-space interfaces, introducing game-specific biases and irrelevant dynamics that hinder world-model training. To address this problem, we introduce GameUI-Taxonomy and G2WEngine, a full-stack framework that formalizes gameplay UI grounding and removal. G2WEngine automatically extracts reusable UI assets from real gameplay videos and synthesizes temporally co
副分类：multimodal
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-08-26-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记 ABot-World-0 精读与批判 — 单桌面 GPU 实时长视野交互世界模型(LongForcing 因果蒸馏 + 多源数据基础设施)

笔记精读 · Perception-R1（Visual Perception Reward for MLLM RLVR）

笔记 Cameron R. Wolfe (Deep Learning Focus) · RSS 摘要

笔记 Cameron Wolfe Substack 短评（双稿）· Agentic World Models + Agentic RL — flyP

仓库 prakash-kharvi/metaverse-screening-record

笔记精读与批判 · MiniWorld: Democratizing the Training of Video World Models from Scratch

arXiv

2608.24680

TLDR

Video games provide a scalable source of training data for video world models, offering diverse environments, complex interactions, and abundant in-the-wild gameplay videos. However, raw gameplay footage entangles the game world with screen-space interfaces, introducing game-specific biases and irrelevant dynamics that hinder world-model training. To address this problem, we introduce GameUI-Taxonomy and G2WEngine, a full-stack framework that formalizes gameplay UI grounding and removal. G2WEngine automatically extracts reusable UI assets from real gameplay videos and synthesizes temporally co

arXiv ↗ 中文全文