arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2307.04721cs.AIcs.CLcs.RO
收藏

大型语言模型作为通用模式机

Large Language Models as General Pattern Machines

Stanford University(斯坦福大学);Google DeepMind(谷歌DeepMind);TU Berlin(柏林工业大学)

Suvir Mirchandani, Fei Xia, Pete Florence, Brian Ichter, Danny Driess, Montserrat Gonzalez Arenas, Kanishka Rao, Dorsa Sadigh, Andy Zeng

AI总结:

本文发现预训练大型语言模型无需额外训练即可作为通用序列建模器,补全复杂 token 序列和空间模式,并探索了其在机器人低级控制中的零样本应用潜力。

摘要:

我们观察到,预训练的大型语言模型(LLMs)能够自回归地补全复杂的 token 序列——从由概率上下文无关文法(PCFG)程序化生成的任意序列,到在通用 AI 基准 Abstraction and Reasoning Corpus(ARC)中发现的更丰富的空间模式,这些模式以 ASCII 艺术风格进行提示。令人惊讶的是,即使序列使用从词汇表中随机采样的 token 来表示,模式补全能力也能部分保留。这些结果表明,无需任何额外训练,LLMs 就可以作为通用序列建模器,由上下文学习驱动。在这项工作中,我们研究了这些零样本能力如何应用于机器人学中的问题——从外推表示随时间变化状态的数字序列以完成简单运动,到对奖励条件轨迹进行 least-to-most 提示,从而发现并表示闭环策略(例如,CartPole 的稳定控制器)。尽管由于延迟、上下文大小限制和计算成本,目前难以部署到真实系统中,但使用 LLMs 驱动低级控制的方法可能提供了一个令人兴奋的视角,让我们看到词语之间的模式如何能够迁移到动作上。

英文摘要:

We observe that pre-trained large language models (LLMs) are capable of autoregressively completing complex token sequences -- from arbitrary ones procedurally generated by probabilistic context-free grammars (PCFG), to more rich spatial patterns found in the Abstraction and Reasoning Corpus (ARC), a general AI benchmark, prompted in the style of ASCII art. Surprisingly, pattern completion proficiency can be partially retained even when the sequences are expressed using tokens randomly sampled from the vocabulary. These results suggest that without any additional training, LLMs can serve as general sequence modelers, driven by in-context learning. In this work, we investigate how these zero-shot capabilities may be applied to problems in robotics -- from extrapolating sequences of numbers that represent states over time to complete simple motions, to least-to-most prompting of reward-conditioned trajectories that can discover and represent closed-loop policies (e.g., a stabilizing controller for CartPole). While difficult to deploy today for real systems due to latency, context size limitations, and compute costs, the approach of using LLMs to drive low-level control may provide an exciting glimpse into how the patterns among words could be transferred to actions.

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。

↑