Game2Player scales visual agent gameplay with co-developed web games
Game2Player: Scaling Gameplay Experience for Visual Agents through Environment-Player Co-Development
Computer Vision and Pattern Recognition
Summary
Playing video games is hard for computer programs because they need to see well, plan ahead, and control actions accurately. Collecting real human gameplay data is expensive, and watching game videos doesn't always show what actions were taken. The authors created Game2Player, a setup where a Builder program modifies games and sets tasks, and Player programs try to complete them while giving feedback to improve the games. This process gathered a large, verified collection of game-playing experiences that helped train better visual agents to play games more effectively.
What this means in practice
- •For video game ai developers: Use the Game2Player framework to create many verified web games and tasks for training and testing vision-based game-playing agents.
- •For interactive media designers: Develop adaptive game environments that are automatically tested and improved by AI players providing feedback for better player experience.
Authors
Xirui Jiang, Yuxi Chen, Rui Yang, Feng Luo, Pengcheng Wang, Bingxuan Li, Zihao Zhao, Yushan Li, Han Wang, Manling Li, Yang Zhang, Shiyu Chang, Tong Zhang, Huan Zhang
Abstract
Games are a rich testbed for visual agents, demanding perception, long-horizon planning, and precise control. However, collecting human gameplay is costly, and learning from gameplay videos requires inferring missing actions. Existing datasets also cover a limited range of games, often without explicit tasks or verifiable outcomes. To address these gaps, we introduce Game2Player, a framework that co-develops game environments and players to collect diverse and verifiable gameplay experience at scale. In the Game2Player loop, a Builder agent adapts existing web games, exposes their internal state, and defines tasks with state-based progress measures and success checks. A Coding Player and a VLM Player then attempt the tasks and report problems to the Builder for repair, so that play serves as both environment validation and training data collection. With Game2Player, we build 568 verified web games with 3,436 tasks, collect 15.7K trajectories (307 hours of gameplay), and hold out 50 games as the benchmark PlayerBench. To evaluate the collected experience as training data, we train VLM agents with progress-weighted SFT, assigning greater weight to steps that advance task progress. Game2Player-9B improves progress from 14.3% to 31.2% on PlayerBench and from 30.0% to 53.1% on the external GameWorld benchmark. More broadly, our work suggests a path for visual agents to scale their gameplay experience by building, verifying, and learning from existing game worlds.