Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912673575206912 |
|---|---|
| author | Wang, Zihao Li, Xujing Ye, Yining Fang, Junjie Wang, Haoming Liu, Longxiang Liang, Shihao Lu, Junting Wu, Zhiyong Feng, Jiazhan Zhong, Wanjun Li, Zili Wang, Yu Miao, Yu Zhou, Bo Li, Yuanfan Wang, Hao Zhao, Zhongkai Wu, Faming Jiang, Zhengxuan Tan, Weihao Yao, Heyuan Yan, Shi Li, Xiangyang Liang, Yitao Qin, Yujia Shi, Guang |
| author_facet | Wang, Zihao Li, Xujing Ye, Yining Fang, Junjie Wang, Haoming Liu, Longxiang Liang, Shihao Lu, Junting Wu, Zhiyong Feng, Jiazhan Zhong, Wanjun Li, Zili Wang, Yu Miao, Yu Zhou, Bo Li, Yuanfan Wang, Hao Zhao, Zhongkai Wu, Faming Jiang, Zhengxuan Tan, Weihao Yao, Heyuan Yan, Shi Li, Xiangyang Liang, Yitao Qin, Yujia Shi, Guang |
| contents | We present Game-TARS, a generalist game agent trained with a unified, scalable action space anchored to human-aligned native keyboard-mouse inputs. Unlike API- or GUI-based approaches, this paradigm enables large-scale continual pre-training across heterogeneous domains, including OS, web, and simulation games. Game-TARS is pre-trained on over 500B tokens with diverse trajectories and multimodal data. Key techniques include a decaying continual loss to reduce causal confusion and an efficient Sparse-Thinking strategy that balances reasoning depth and inference cost. Experiments show that Game-TARS achieves about 2 times the success rate over the previous sota model on open-world Minecraft tasks, is close to the generality of fresh humans in unseen web 3d games, and outperforms GPT-5, Gemini-2.5-Pro, and Claude-4-Sonnet in FPS benchmarks. Scaling results on training-time and test-time confirm that the unified action space sustains improvements when scaled to cross-game and multimodal data. Our results demonstrate that simple, scalable action representations combined with large-scale pre-training provide a promising path toward generalist agents with broad computer-use abilities. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_23691 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents Wang, Zihao Li, Xujing Ye, Yining Fang, Junjie Wang, Haoming Liu, Longxiang Liang, Shihao Lu, Junting Wu, Zhiyong Feng, Jiazhan Zhong, Wanjun Li, Zili Wang, Yu Miao, Yu Zhou, Bo Li, Yuanfan Wang, Hao Zhao, Zhongkai Wu, Faming Jiang, Zhengxuan Tan, Weihao Yao, Heyuan Yan, Shi Li, Xiangyang Liang, Yitao Qin, Yujia Shi, Guang Artificial Intelligence We present Game-TARS, a generalist game agent trained with a unified, scalable action space anchored to human-aligned native keyboard-mouse inputs. Unlike API- or GUI-based approaches, this paradigm enables large-scale continual pre-training across heterogeneous domains, including OS, web, and simulation games. Game-TARS is pre-trained on over 500B tokens with diverse trajectories and multimodal data. Key techniques include a decaying continual loss to reduce causal confusion and an efficient Sparse-Thinking strategy that balances reasoning depth and inference cost. Experiments show that Game-TARS achieves about 2 times the success rate over the previous sota model on open-world Minecraft tasks, is close to the generality of fresh humans in unseen web 3d games, and outperforms GPT-5, Gemini-2.5-Pro, and Claude-4-Sonnet in FPS benchmarks. Scaling results on training-time and test-time confirm that the unified action space sustains improvements when scaled to cross-game and multimodal data. Our results demonstrate that simple, scalable action representations combined with large-scale pre-training provide a promising path toward generalist agents with broad computer-use abilities. |
| title | Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2510.23691 |