GTAGym, 게임 기반 에이전트 평가·강화학습 환경 제시
GTAGym: Real AAA Game Tasks Environment for Evaluation and Reinforcement Learning
- 매체
- Hugging Face
- 게시
- 중요도
- 2/5
- 화제 지수
- 10
한국어 요약
AAA 게임 환경을 에이전트 평가와 강화학습 테스트 플랫폼으로 활용하기 위해 GTAGym과 GTAVBench를 제시했다. GTA V 메인 스토리 미션을 기반으로 103개 온라인 과제와 47개 슈팅 과제를 구성했고, 웹 서버를 통해 자동 결과 확인이 가능하다. 최고 구성에서 GTAVBench-AGI에서 70.9% 정확도를 기록했고, Qwen3.8-27B LoRA 강화학습은 도메인 내·외 과제 모두에서 성능을 15%에서 23%로 향상시켰다.
원문 발췌 · English
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%.