VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks
Summary
VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks (2024) extends web-agent evaluation from text-only pages to tasks that need visual information. Most computer interfaces are built for human eyes, so many realistic tasks cannot be solved from HTML or accessibility trees alone.
What it introduced: a benchmark of visually grounded web tasks in which an agent must read image-text inputs, follow natural-language instructions, and act on websites. The authors compare text-only and multimodal agents and document gaps in both: text-only models miss visual cues, and even strong multimodal agents still fail many tasks.
What it led to: a standard multimodal companion to WebArena, used when people ask whether a web agent can handle screenshots, icons, and other visual page structure. Live notes: 2307-13854-webarena. Mind2Web and WebVoyager are pointer notes in this batch; WebShop is a full note in batch F.
Where to read it: https://arxiv.org/abs/2401.13649 (v2, last revised 6 Jun 2024). Accepted to ACL 2024. Project page: https://jykoh.com/vwa.
The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.
License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.
License: CC-BY-4.0 · Pointer by iinventedlight (AI-assisted). Source: Jing Yu Koh, Robert Lo, Lawrence Jang, and 7 coauthors, VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks (arXiv non-exclusive distribution license). This note links to the paper rather than reproducing it.
Cite this
'iinventedlight (AI-assisted)'. 2026. "VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks." iinventedlight Library, iil:note:4afa6akj, version of 2026-11-12. https://iinventedlight.com/library/notes/2401-13649-visualwebarena/. License: CC-BY-4.0.
@misc{iil_note_4afa6akj,
author = {{'iinventedlight (AI-assisted)'}},
title = {{VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks}},
howpublished = {iinventedlight Library},
year = {2026},
date = {2026-11-12},
url = {https://iinventedlight.com/library/notes/2401-13649-visualwebarena/},
note = {Permanent ID iil:note:4afa6akj. License: CC-BY-4.0.},
}
Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.
For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2401-13649-visualwebarena.md.