WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?

TL;DR

This paper introduces WorkArena, a benchmark for evaluating large language model-based web agents on enterprise software tasks, revealing current limitations and disparities between open and closed-source models.

Contribution

We present WorkArena, a new benchmark with 33 enterprise tasks and BrowserGym environment for evaluating web agents, highlighting performance gaps and areas for improvement.

Findings

Current agents perform variably on WorkArena tasks.

Significant performance gap exists between open and closed-source LLMs.

Full task automation remains a challenge for current agents.

Abstract

We study the use of large language model-based agents for interacting with software via web browsers. Unlike prior work, we focus on measuring the agents' ability to perform tasks that span the typical daily work of knowledge workers utilizing enterprise software systems. To this end, we propose WorkArena, a remote-hosted benchmark of 33 tasks based on the widely-used ServiceNow platform. We also introduce BrowserGym, an environment for the design and evaluation of such agents, offering a rich set of actions as well as multimodal observations. Our empirical evaluation reveals that while current agents show promise on WorkArena, there remains a considerable gap towards achieving full task automation. Notably, our analysis uncovers a significant performance disparity between open and closed-source LLMs, highlighting a critical area for future exploration and development in the field.