arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.00690cs.AI
收藏

使用 LLM 玩 NetHack:作为零样本智能体的潜力与局限

Playing NetHack with LLMs: Potential & Limitations as Zero-Shot Agents

Queen Mary University of London(伦敦玛丽女王大学);Creative Assembly(创意装配公司)

Dominik Jeurissen, Diego Perez-Liebana, Jeremy Gow, Duygu Cakmak, James Kwan

AI总结:

本文提出首个面向 NetHack 的 LLM 零样本智能体 NetPlay,通过技能选择、历史追踪与事件中断机制验证其潜力,并指出详细动态上下文对复杂游戏决策的必要性。

摘要:

大语言模型(LLM)作为零样本游戏智能体的高层规划器已展现出巨大成功。然而,这些智能体主要在 Minecraft 上接受评估,而该游戏中的长期规划相对直接。相比之下,在动态机器人环境中测试的智能体则因环境过于简单、仅有少量物体和交互而面临局限。为填补文献中的这一空白,我们提出了 NetPlay,这是首个由 LLM 驱动、面向具有挑战性的 roguelike 游戏 NetHack 的零样本智能体。NetHack 因其多样的物品与怪物、复杂的交互以及众多死亡方式而成为一个极具挑战性的环境。NetPlay 采用一种为动态机器人环境设计的架构,并针对 NetHack 进行了修改。与以往方法类似,它提示 LLM 从预定义技能中进行选择,并追踪过去的交互以增强决策。鉴于 NetHack 不可预测的特性,NetPlay 会检测重要游戏事件以中断正在运行的技能,从而使其能够对意外情况作出反应。尽管 NetPlay 在与 NetHack 的机制交互方面表现出相当强的灵活性和熟练度,但它在处理模糊任务描述和缺乏明确反馈时存在困难。我们的研究结果表明,NetPlay 在具备详细上下文信息时表现最佳,这说明在 NetHack 等复杂游戏中,需要采用动态方法来提供上下文信息。

英文摘要:

Large Language Models (LLMs) have shown great success as high-level planners for zero-shot game-playing agents. However, these agents are primarily evaluated on Minecraft, where long-term planning is relatively straightforward. In contrast, agents tested in dynamic robot environments face limitations due to simplistic environments with only a few objects and interactions. To fill this gap in the literature, we present NetPlay, the first LLM-powered zero-shot agent for the challenging roguelike NetHack. NetHack is a particularly challenging environment due to its diverse set of items and monsters, complex interactions, and many ways to die. NetPlay uses an architecture designed for dynamic robot environments, modified for NetHack. Like previous approaches, it prompts the LLM to choose from predefined skills and tracks past interactions to enhance decision-making. Given NetHack's unpredictable nature, NetPlay detects important game events to interrupt running skills, enabling it to react to unforeseen circumstances. While NetPlay demonstrates considerable flexibility and proficiency in interacting with NetHack's mechanics, it struggles with ambiguous task descriptions and a lack of explicit feedback. Our findings demonstrate that NetPlay performs best with detailed context information, indicating the necessity for dynamic methods in supplying context information for complex games such as NetHack.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑