SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2401.13649 · 2024

VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 failures of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2401.13649.json · badge: /api/ran/2401.13649.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
web-arena-x/visualwebarena canonical (classifier label)
classifier basis: census
7 ran · 0 no recorded run
FunctionStatusWhere it lives
action2str Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/actions.py
code served (permissive licence) · get_code("32bc38b0f6c79438")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import numpy.typing as npt

# Define the _id2key mapping that the function uses
_id2key = {
    0: 'a', 1: 'b', 2: 'c', 3: 'd', 4: 'e', 5: 'f', 6: 'g', 7: 'h', 8: 'i', 9: 'j',
    10: 'k', 11: 'l', 12: 'm', 13: 'n', 14: 'o', 15: 'p', 16: 'q', 17: 'r', 18: 's', 19: 't',
    20: 'u', 21: 'v', 22: 'w', 23: 'x', 24: 'y', 25: 'z'
}

# Create a valid Action dict for CLICK action type
action_click = {
    'action_type': 6,  # ActionTypes.CLICK
    'coords': np.array([100.5, 200.5], dtype=np.float32),
    'element_role': 1,
    'element_name': 'button',
    'text': [0, 1, 2],  # indices into _id2key
    'page_number': 1,
    'url': 'https://example.com',
    'nth': 0,
    'element_id': 'elem_123',
    'direction': 'down',
    'key_comb': 'ctrl+c',
    'pw_code': 'pwd123',
    'answer': 'yes',
    'raw_prediction': 'raw_pred'
}

ARGS = {
    'action': action_click,
    'action_set_tag': 'id_accessibility_tree',
    'semantic_element': 'A clickable button element'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
beartype, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_site_comb_from_filepath Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/auto_login.py
code served (permissive licence) · get_code("824ae078c5d3295a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os

def get_site_comb_from_filepath(file_path: str) -> list[str]:
    comb = os.path.basename(file_path).rsplit('_', 1)[0].split('.')
    return comb

# The function takes a file_path string and extracts the basename,
# splits by '_' from the right, takes the first part, then splits by '.'
# Example: "/path/to/file.site1.site2_suffix.txt" -> basename -> "file.site1.site2_suffix.txt"
# -> rsplit('_', 1)[0] -> "file.site1.site2_suffix" -> split('.') -> ["file", "site1", "site2_suffix"]

ARGS = {
    "file_path": "/path/to/file.site1.site2_suffix.txt"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_action Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/envs.py
code served (permissive licence) · get_code("1829c408a278c974")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned PlaywrightScript. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from dataclasses import dataclass

@dataclass
class PlaywrightScript:
    function: str
    destination: str
    name: str | None = None
    operation: str | None = None
    value: str | None = None

def parse_action(action: str) -> PlaywrightScript:
    splitted = action.strip().split(' ')
    assert len(splitted) >= 2
    match splitted[:2]:
        case ['goto', url]:
            assert len(splitted) == 2
            return PlaywrightScript('goto', url)
        case ['get_by_role', destination]:
            assert len(splitted) >= 4
            match splitted[2:]:
                case [name, operation]:
                    return PlaywrightScript('get_by_role', destination, name, operation)
                case [name, operation, value]:
                    return PlaywrightScript('get_by_role', destination, name, operation, value)
                case _:
                    raise ValueError('Invalid action')
        case _:
            raise ValueError(f'Invalid action {action}')

ARGS = {
    'action': 'goto https://example.com'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PlaywrightScript — a PlaywrightScript output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PlaywrightScript output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
pil_to_b64 Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from osu-nlp-group/webdreamer, is pointer only (licence: NONE) · get_code("d8254925e6697a59")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from osu-nlp-group/webdreamer, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

pil_to_vertex Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from osu-nlp-group/webdreamer, is pointer only (licence: NONE) · get_code("71ec370bd2568e5b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from osu-nlp-group/webdreamer, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

png_bytes_to_numpy Ran
ran, honoured the contract we drafted
web-arena-x/visualwebarena/browser_env/utils.py
code served (permissive licence) · get_code("04ba3dcf1bb13e4f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family int, shape (4, 3, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from PIL import Image
from io import BytesIO

# Create a simple PNG image in bytes format
# Create a small 4x3 RGB image (distinct dimensions for clarity)
img_array = np.random.randint(0, 256, (4, 3, 3), dtype=np.uint8)
img = Image.fromarray(img_array, mode='RGB')

# Convert to PNG bytes
png_buffer = BytesIO()
img.save(png_buffer, format='PNG')
png_bytes = png_buffer.getvalue()

ARGS = {
    'png': png_bytes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, beartype, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (4, 3, 3) — shape (4, 3, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
remove_unicode Ran
ran; our contract draft was wrong, not the code
web-arena-x/visualwebarena/browser_env/processors.py
code served (permissive licence) · get_code("b3cfdfe328f76c2d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def remove_unicode(input_string):
    unicode_pattern = re.compile('[^\\x00-\\x7F]+')
    cleaned_string = unicode_pattern.sub('', input_string)
    return cleaned_string

ARGS = {
    'input_string': 'Hello World! こんにちは 你好 🌍'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 6 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 6 later) and 1 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
get_site_comb_from_filepath 4 other papers
2 dated earlier, 2 later
2305.14257 (2023-05)
Hierarchical Prompting Assists Large Language Model on Web Navigation
2307.13854 (2023-07)
WebArena: A Realistic Web Environment for Building Autonomous Agents
2407.01476 (2024-07)
Tree Search for Language Model Agents
and 1 more not named here
parse_action 4 other papers
2 dated earlier, 2 later
2305.14257 (2023-05)
Hierarchical Prompting Assists Large Language Model on Web Navigation
2307.13854 (2023-07)
WebArena: A Realistic Web Environment for Building Autonomous Agents
2407.01476 (2024-07)
Tree Search for Language Model Agents
and 1 more not named here
pil_to_b64 3 other papers
3 later
2407.01476 (2024-07)
Tree Search for Language Model Agents
2411.06559 (2024-11)
Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents
2505.16421 (2025-05)
WEBAGENT-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning
pil_to_vertex 3 other papers
3 later
2407.01476 (2024-07)
Tree Search for Language Model Agents
2411.06559 (2024-11)
Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents
2505.16421 (2025-05)
WEBAGENT-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning
png_bytes_to_numpy 2 other papers
2 later
2407.01476 (2024-07)
Tree Search for Language Model Agents
2505.16421 (2025-05)
WEBAGENT-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning
remove_unicode 2 other papers
2 later
2407.01476 (2024-07)
Tree Search for Language Model Agents
2505.16421 (2025-05)
WEBAGENT-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Autonomous agents capable of planning, reasoning, and executing actions on the web offer a promising avenue for automating computer tasks. However, the majority of existing benchmarks primarily focus on text-based agents, neglecting many natural tasks that require visual information to effectively solve. Given that most computer interfaces cater to human perception, visual information often augments textual data in ways that text-only models struggle to harness effectively. To bridge this gap, we introduce VisualWebArena, a benchmark designed to assess the performance of multimodal web agents on realistic \textit{visually grounded tasks}. VisualWebArena comprises of a set of diverse and complex web-based tasks that evaluate various capabilities of autonomous multimodal agents. To perform on this benchmark, agents need to accurately process image-text inputs, interpret natural language instructions, and execute actions on websites to accomplish user-defined objectives. We conduct an extensive evaluation of state-of-the-art LLM-based autonomous agents, including several multimodal models. Through extensive quantitative and qualitative analysis, we identify several limitations of text-only LLM agents, and reveal gaps in the capabilities of state-of-the-art multimodal language agents. VisualWebArena provides a framework for evaluating multimodal autonomous language agents, and offers insights towards building stronger autonomous agents for the web. Our code, baseline models, and data is publicly available at https://jykoh.com/vwa.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2401.13649")
get_code_for_paper("2401.13649")
have("2401.13649")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2401.13649.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2401.13649.svg)](https://syntology.ai/paper/2401.13649)

Connect an agent — have() is free.