arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.08140cs.CL
收藏

BAGEL:通过语言引导探索来引导智能体

BAGEL: Bootstrapping Agents by Guiding Exploration with Language

Stanford University(斯坦福大学);Google Deepmind(谷歌DeepMind)

Shikhar Murty, Christopher Manning, Peter Shaw, Mandar Joshi, Kenton Lee

AI总结:

BAGEL 通过语言标注器与零样本智能体的迭代往返,将随机轨迹自动转化为自然语言演示,无需人类监督即可提升 LM 智能体在新环境中的泛化能力,并在 ToolQA 和 MiniWob++ 上取得显著性能提升。

摘要:

在数字环境(如网页浏览器和 REST API)中通过执行动作来遵循自然语言指令,对语言模型(LM)智能体而言是一项具有挑战性的任务。遗憾的是,在没有人类演示的情况下,LM 智能体往往难以泛化到新环境。本文提出了 BAGEL,一种无需人类监督即可引导 LM 智能体的方法。BAGEL 通过两个带噪声的 LM 组件之间的往返过程,将随机探索轨迹或合成指令的种子集转化为演示:一个 LM 标注器将轨迹转化为合成指令,一个零样本 LM 智能体将合成指令映射为精化后的轨迹。通过迭代执行这些往返过程,BAGEL 能迅速将初始轨迹分布转变为那些能被自然语言良好描述的轨迹。我们利用 BAGEL 演示,在测试时通过基于检索演示的上下文学习来适配零样本 LM 智能体,并在 ToolQA 和 MiniWob++ 上发现绝对性能提升超过 2-13%,执行失败次数最多减少 13 倍。

英文摘要:

Following natural language instructions by executing actions in digital environments (e.g. web-browsers and REST APIs) is a challenging task for language model (LM) agents. Unfortunately, LM agents often fail to generalize to new environments without human demonstrations. This work presents BAGEL, a method for bootstrapping LM agents without human supervision. BAGEL converts a seed set of randomly explored trajectories or synthetic instructions, into demonstrations, via round-trips between two noisy LM components: an LM labeler which converts a trajectory into a synthetic instruction, and a zero-shot LM agent which maps the synthetic instruction into a refined trajectory. By performing these round-trips iteratively, BAGEL quickly converts the initial distribution of trajectories towards those that are well-described by natural language. We use BAGEL demonstrations to adapt a zero shot LM agent at test time via in-context learning over retrieved demonstrations, and find improvements of over 2-13% absolute on ToolQA and MiniWob++, with up to 13x reduction in execution failures.

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。

↑