A Zero-Shot Language Agent for Computer Control with Structured Reflection

TL;DR

This paper introduces a zero-shot language-based agent capable of controlling a computer environment by planning, reflecting, and learning from its mistakes without expert traces, achieving competitive performance on MiniWoB++ tasks.

Contribution

The paper presents a novel zero-shot agent that uses structured reflection and self-improvement to perform computer control tasks without relying on trace examples or prior training.

Findings

Outperforms recent methods on simple MiniWoB++ tasks

Performs comparably to state-of-the-art on complex tasks

Operates effectively without expert demonstrations or additional screen info

Abstract

Large language models (LLMs) have shown increasing capacity at planning and executing a high-level goal in a live computer environment (e.g. MiniWoB++). To perform a task, recent works often require a model to learn from trace examples of the task via either supervised learning or few/many-shot prompting. Without these trace examples, it remains a challenge how an agent can autonomously learn and improve its control on a computer, which limits the ability of an agent to perform a new task. We approach this problem with a zero-shot agent that requires no given expert traces. Our agent plans for executable actions on a partially observed environment, and iteratively progresses a task by identifying and learning from its mistakes via self-reflection and structured thought management. On the easy tasks of MiniWoB++, we show that our zero-shot agent often outperforms recent SoTAs, with more efficient reasoning. For tasks with more complexity, our reflective agent performs on par with prior best models, even though previous works had the advantages of accessing expert traces or additional screen information.