SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.08264 · 2024

Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 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, 1 violated it and 0 have no contract verdict; 4 failures of our instrument; 2 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/2409.08264.json · badge: /api/ran/2409.08264.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
microsoft/windowsagentarena canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
get_directory_structure Ran
ran; our contract draft was wrong, not the code
microsoft/windowsagentarena/src/win-arena-container/client/evaluation_examples_windows/create_json.py
code served (permissive licence) · get_code("e95e71722d8ee6d7")
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 tuple, 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
import tempfile
import shutil

# Create a temporary directory structure for testing
temp_base_dir = tempfile.mkdtemp()

# Create some subdirectories with files
subdir1 = os.path.join(temp_base_dir, "category_A")
subdir2 = os.path.join(temp_base_dir, "category_B")
subdir3 = os.path.join(temp_base_dir, "exclude_me")

os.makedirs(subdir1, exist_ok=True)
os.makedirs(subdir2, exist_ok=True)
os.makedirs(subdir3, exist_ok=True)

# Create some dummy files
open(os.path.join(subdir1, "file1.txt"), 'w').close()
open(os.path.join(subdir1, "file2.txt"), 'w').close()
open(os.path.join(subdir2, "file3.txt"), 'w').close()
open(os.path.join(subdir3, "file4.txt"), 'w').close()

ARGS = {
    "base_dir": temp_base_dir,
    "filter_out_categories": ["exclude"]
}
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 tuple, ndim 1, dtype family float — a tuple 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
get_result Ran
ran, honoured the contract we drafted
microsoft/windowsagentarena/src/win-arena-container/client/show_result.py
code served (permissive licence) · get_code("c84f9385b9ff870c")
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
import json
import tempfile
import shutil

# Create a temporary directory structure that mimics the expected format
temp_dir = tempfile.mkdtemp()

# Create domain directories with example subdirectories and result.txt files
domains = ['libreoffice_calc', 'libreoffice_impress', 'libreoffice_writer', 
           'vlc', 'thunderbird', 'chrome', 'gimp', 'vs_code']

for domain in domains:
    domain_path = os.path.join(temp_dir, domain)
    os.makedirs(domain_path, exist_ok=True)
    
    # Create a few example directories with result.txt files
    for example_id in ['example_1', 'example_2', 'example_3']:
        example_path = os.path.join(domain_path, example_id)
        os.makedirs(example_path, exist_ok=True)
        
        # Create result.txt with a float value (0.0 to 1.0 representing success/failure)
        result_value = 0.8 if example_id == 'example_1' else 0.5
        with open(os.path.join(example_path, 'result.txt'), 'w') as f:
            f.write(str(result_value))

# Add one example with missing result.txt to test that path
special_domain_path = os.path.join(temp_dir, 'chrome')
special_example_path = os.path.join(special_domain_path, 'example_4')
os.makedirs(special_example_path, exist_ok=True)
# Don't create result.txt for this one

ARGS = {
    'target_dir': temp_dir
}
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
is_leaf_node Ran
ran, violated the contract we drafted
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/a11y_demo.py
code served (permissive licence) · get_code("a3a1e6059333e000")
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 bool, ndim 0, dtype family float, shape () 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 torch

torch.manual_seed(0)

class MockNode:
    def __init__(self, children=None):
        self.children = children if children is not None else []
    
    def __iter__(self):
        return iter(self.children)

# Test case 1: leaf node (no children)
leaf_node = MockNode(children=[])

# Test case 2: non-leaf node (has children)
non_leaf_node = MockNode(children=[MockNode(), MockNode()])

ARGS = {
    "node": leaf_node
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () 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
prev_actions_to_string Ran
ran; our contract draft was wrong, not the code
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/agent.py
code served (permissive licence) · get_code("2a014475a5e7301b")
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 torch

torch.manual_seed(0)

# prev_actions should be a list of strings representing previous actions
prev_actions = [
    "Action 1: moved mouse to (100, 200)",
    "Action 2: clicked button",
    "Action 3: typed 'hello'",
    "Action 4: pressed Enter",
    "Action 5: scrolled down"
]

n_prev = 3

ARGS = {
    "prev_actions": prev_actions,
    "n_prev": n_prev
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
remove_min_leading_spaces Ran
ran; our contract draft was wrong, not the code
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/agent.py
code served (permissive licence) · get_code("06c42a797d1bb70c")
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 torch
import numpy as np

# Analyzing the function:
# - It takes a single parameter: text (a string)
# - It splits by newlines, finds minimum leading spaces, and removes that amount from each line
# - The function works with string manipulation only

text = """    hello world
    foo bar
  baz"""

ARGS = {
    "text": text
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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
resize_image_openai Ran
ran; our contract draft was wrong, not the code
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/agent.py
code served (permissive licence) · get_code("7aa29e3ad527ce33")
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 tuple. 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 PIL import Image

# Create a test image using PIL
# The function expects an image object with .size attribute and .resize() method
# This matches PIL.Image.Image interface

# Create a large image that will trigger resizing
test_image = Image.new('RGB', (3000, 2500), color='red')

ARGS = {
    'image': test_image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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_a11y Not yet run
unverified (harvested, no recorded run)
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/a11y_demo.py
code served (permissive licence) · get_code("1e0a1c4c6aeb27c3")
How it ran

no invocation recorded.

get_screen Not yet run
unverified (harvested, no recorded run)
microsoft/windowsagentarena/src/win-arena-container/client/mm_agents/navi/a11y_demo.py
code served (permissive licence) · get_code("42369f375e5d6c37")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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

Large language models (LLMs) show remarkable potential to act as computer agents, enhancing human productivity and software accessibility in multi-modal tasks that require planning and reasoning. However, measuring agent performance in realistic environments remains a challenge since: (i) most benchmarks are limited to specific modalities or domains (e.g. text-only, web navigation, Q&A, coding) and (ii) full benchmark evaluations are slow (on order of magnitude of days) given the multi-step sequential nature of tasks. To address these challenges, we introduce the Windows Agent Arena: a reproducible, general environment focusing exclusively on the Windows operating system (OS) where agents can operate freely within a real Windows OS and use the same wide range of applications, tools, and web browsers available to human users when solving tasks. We adapt the OSWorld framework (Xie et al., 2024) to create 150+ diverse Windows tasks across representative domains that require agent abilities in planning, screen understanding, and tool usage. Our benchmark is scalable and can be seamlessly parallelized in Azure for a full benchmark evaluation in as little as 20 minutes. To demonstrate Windows Agent Arena's capabilities, we also introduce a new multi-modal agent, Navi. Our agent achieves a success rate of 19.5% in the Windows domain, compared to 74.5% performance of an unassisted human. Navi also demonstrates strong performance on another popular web-based benchmark, Mind2Web. We offer extensive quantitative and qualitative analysis of Navi's performance, and provide insights into the opportunities for future research in agent development and data generation using Windows Agent Arena. Webpage: https://microsoft.github.io/WindowsAgentArena Code: https://github.com/microsoft/WindowsAgentArena

For agents

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

get_harvested_code_for_paper("2409.08264")
get_code_for_paper("2409.08264")
have("2409.08264")

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

curl https://syntology.ai/api/ran/2409.08264.json

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

[![Syntology run record](https://syntology.ai/api/ran/2409.08264.svg)](https://syntology.ai/paper/2409.08264)

Connect an agent — have() is free.