arXiv cs.CL· Jiajun Chen, Shicheng Fang, Jiahao Lu, Yuxin Wang, Xipeng Qiu·· 10 小时前AI 评分42
GTAGym：面向 GTA V 任务的评测与强化学习 AAA 游戏环境
GTAGym: Real AAA Game Tasks Environment for Evaluation and Reinforcement Learning
AI 导读
GTAGym 是面向 GTA V 任务的评测与强化学习环境，其基准 GTAVBench 包含 103 个在线任务（覆盖九个任务族）和 47 个基于 GTA V 图像的射击狙击任务。
正文 · 原文
Abstract:AAA games offer a rich testbed for evaluating agent perception, reasoning, and control. However, evaluating whether agents can complete tasks in AAA games is challenging and often human-annotated because task outcomes are often difficult to verify automatically, let alone applying it for Reinforcement Learning (RL). We introduce GTAGym, an environment and framework for evaluation and RL on GTA V tasks. Its evaluation benchmark, GTAVBench, derived from GTA V main-story missions, contains 103 online tasks reconstructed across nine task families, alongside 47 image-based shooting and sniping tasks using GTA V imagery. A web server provides access to the online environment and evaluation interface, supporting parallel testing through screenshots and keyboard-mouse actions with automatic engine-state outcome checks. We evaluate six frontier LLMs on GTAVBench with textual task instructions and on its GTAVBench-AGI branch, where instructions appear only in the first screenshot and a harness automatically compresses interaction history. The best configuration achieves 70.9% overall accuracy in GTAVBench-AGI, yet aerial navigation, evasion and interception remain difficult. GTAGym also provides a scalable generator that samples task parameters within user-specified ranges and option sets under fixed templates, creating batches of 3D and image-based shooting environments for RL with automatic reward and termination signals. LoRA RL of Qwen3.8-27B on generated on-foot navigation tasks improves the in-domain performance and transfers to out-of-domain task categories, and native GTA V tasks from 15% to 23%.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.13028 [cs.CV] |
| (or arXiv:2611.13028v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13028 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jiajun Chen [view email]
[v1]
Mon, 9 Nov 2026 08:59:36 UTC (3,330 KB)
来源：arXiv cs.CL · arxiv.org