CLAIHCLGMar 30, 2023

Language Models can Solve Computer Tasks

arXiv:2303.17491v3506 citationsh-index: 119Has Code
Originality Highly original
AI Analysis

This work addresses the challenge of creating practical agents for general computer task automation, reducing reliance on large expert datasets and task-specific rewards, which is significant for improving efficiency in productivity tools.

The paper tackles the problem of automating computer tasks using natural language commands by introducing a prompting scheme called Recursively Criticizes and Improves (RCI) with pre-trained large language models, achieving state-of-the-art results on the MiniWoB++ benchmark and outperforming supervised and reinforcement learning approaches with only a handful of demonstrations per task.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes