Language Models can Solve Computer Tasks

TL;DR

This paper introduces a recursive criticism and improvement (RCI) prompting method for large language models to automate computer tasks using natural language, achieving state-of-the-art results with minimal demonstrations and no task-specific rewards.

Contribution

The paper presents RCI prompting, a novel method enabling LLMs to solve computer tasks efficiently without extensive supervision or reward engineering.

Findings

RCI outperforms existing LLM methods on MiniWoB++

RCI with InstructGPT-3+RLHF achieves state-of-the-art results

RCI enhances reasoning abilities beyond chain of thought prompting

Abstract

Agents capable of carrying out general tasks on a computer can improve efficiency and productivity by automating repetitive tasks and assisting in complex problem-solving. Ideally, such agents should be able to solve new computer tasks presented to them through natural language commands. However, previous approaches to this problem require large amounts of expert demonstrations and task-specific reward functions, both of which are impractical for new tasks. In this work, we show that a pre-trained large language model (LLM) agent can execute computer tasks guided by natural language using a simple prompting scheme where the agent Recursively Criticizes and Improves its output (RCI). The RCI approach significantly outperforms existing LLM methods for automating computer tasks and surpasses supervised learning (SL) and reinforcement learning (RL) approaches on the MiniWoB++ benchmark. We compare multiple LLMs and find that RCI with the InstructGPT-3+RLHF LLM is state-of-the-art on MiniWoB++, using only a handful of demonstrations per task rather than tens of thousands, and without a task-specific reward function. Furthermore, we demonstrate RCI prompting's effectiveness in enhancing LLMs' reasoning abilities on a suite of natural language reasoning tasks, outperforming chain of thought (CoT) prompting with external feedback. We find that RCI combined with CoT performs better than either separately. Our code can be found here: https://github.com/posgnu/rci-agent.