arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2305.02412cs.CLcs.AIcs.LG
收藏

规划、消除与追踪——语言模型是具身智能体的优秀教师

Plan, Eliminate, and Track -- Language Models are Good Teachers for Embodied Agents

Carnegie Mellon University（卡内基梅隆大学）；Ariel University（阿里埃勒大学）；Microsoft Research（微软研究院）；Nvidia Research（英伟达研究院）

Yue Wu, So Yeon Min, Yonatan Bisk, Ruslan Salakhutdinov, Amos Azaria, Yuanzhi Li, Tom Mitchell, Shrimai Prabhumoye

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出PET框架，利用大语言模型将任务分解为子任务、消除无关观察并追踪完成状态，在AlfWorld基准上超越SOTA 15%。

摘要：

预训练的大型语言模型（LLM）捕获了关于世界的程序性知识。近期工作利用LLM生成抽象计划的能力来简化具有挑战性的控制任务，通过动作评分或动作建模（微调）的方式实现。然而，Transformer架构继承了若干限制，使得LLM难以直接充当智能体：例如输入长度有限、微调效率低下、预训练带来的偏差以及与非文本环境的不兼容性。为了保持与低层可训练actor的兼容性，我们提出利用LLM中的知识来简化控制问题，而非直接解决它。我们提出了规划、消除与追踪（PET）框架。规划模块将任务描述转化为高级子任务列表。消除模块从当前子任务的观察中屏蔽掉无关对象和容器。最后，追踪模块判断智能体是否完成了每个子任务。在AlfWorld指令跟随基准上，PET框架在泛化到人类目标规范方面比SOTA显著提升了15%。

英文摘要：

Pre-trained large language models (LLMs) capture procedural knowledge about the world. Recent work has leveraged LLM's ability to generate abstract plans to simplify challenging control tasks, either by action scoring, or action modeling (fine-tuning). However, the transformer architecture inherits several constraints that make it difficult for the LLM to directly serve as the agent: e.g. limited input lengths, fine-tuning inefficiency, bias from pre-training, and incompatibility with non-text environments. To maintain compatibility with a low-level trainable actor, we propose to instead use the knowledge in LLMs to simplify the control problem, rather than solving it. We propose the Plan, Eliminate, and Track (PET) framework. The Plan module translates a task description into a list of high-level sub-tasks. The Eliminate module masks out irrelevant objects and receptacles from the observation for the current sub-task. Finally, the Track module determines whether the agent has accomplished each sub-task. On the AlfWorld instruction following benchmark, the PET framework leads to a significant 15% improvement over SOTA for generalization to human goal specifications.

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。

↑