SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.06327 · 2024

VisualAgentBench: Towards Large Multimodal Models as Visual Foundation Agents

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 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; 2 failures of our instrument; 1 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/2408.06327.json · badge: /api/ran/2408.06327.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
thudm/visualagentbench canonical (classifier label)
classifier basis: census
3 ran · 1 no recorded run
FunctionStatusWhere it lives
analyze_output Ran
ran; our contract draft was wrong, not the code
thudm/visualagentbench/src/analysis.py
code served (permissive licence) · get_code("9a10e3626d5a587e")
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.

import datetime
import json
import os
import tempfile
from collections import OrderedDict

# Create a temporary directory structure for the output
temp_dir = tempfile.mkdtemp()

# Create mock agent and task directories with overall.json files
agent_dir = os.path.join(temp_dir, 'gpt-4o-2024-05-13')
task_dir = os.path.join(agent_dir, 'task1')
os.makedirs(task_dir, exist_ok=True)

# Create a mock overall.json file
overall_data = {
    'COMPLETED': 5,
    'AGENT_CONTEXT_LIMIT': 0,
    'AGENT_VALIDATION_FAILED': 1,
    'AGENT_INVALID_ACTION': 0,
    'UNKNOWN': 0,
    'TASK_ERROR': 0,
    'TASK_LIMIT_REACHED': 0,
    'validation': {
        'COMPLETED': 5,
        'AGENT_CONTEXT_LIMIT': 0,
        'AGENT_VALIDATION_FAILED': 1,
        'AGENT_INVALID_ACTION': 0,
        'UNKNOWN': 0,
        'TASK_ERROR': 0,
        'TASK_LIMIT_REACHED': 0
    }
}

overall_json_path = os.path.join(task_dir, 'overall.json')
with open(overall_json_path, 'w', encoding='utf-8') as f:
    json.dump(overall_data, f)

# Create a mock config file
config_data = {
    'definition': {
        'agent': {
            'gpt-4o-2024-05-13': {},
            'gpt-4': {},
            'llama-2-13b': {}
        },
        'task': {
            'task1': {},
            'task2': {},
            'task3': {}
        }
    }
}

config_file_path = os.path.join(temp_dir, 'config.json')
with open(config_file_path, 'w', encoding='utf-8') as f:
    json.dump(config_data, f)

# Create a minimal ConfigLoader class
class ConfigLoader:
    def load_from(self, path):
        with open(path, 'r', encoding='utf-8') as f:
            return json.load(f)

class ColorMessage:
    @staticmethod
    def cyan(text):
        return text

# Make these available in the global scope for the function to use
import builtins
builtins.ConfigLoader = ConfigLoader
builtins.ColorMessage = ColorMessage

# Define ARGS
ARGS = {
    'config': config_file_path,
    'output': temp_dir,
    'since_timestamp': 0.0
}
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 — 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
deep_merge Ran
ran; our contract draft was wrong, not the code
thudm/visualagentbench/src/configs.py
code served (permissive licence) · get_code("938793bb2e85dded")
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 dict. 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 copy import deepcopy

def deep_merge(base_item, new_item):
    if isinstance(base_item, dict) and isinstance(new_item, dict):
        ret = deepcopy(base_item)
        for key in new_item:
            if key in ret:
                ret[key] = deep_merge(ret[key], new_item[key])
            else:
                ret[key] = new_item[key]
        return ret
    if isinstance(base_item, list) and isinstance(new_item, list):
        ret = deepcopy(base_item)
        ret.extend(new_item)
        return ret
    return new_item


# Construct valid arguments for deep_merge
base_item = {
    "config": {
        "learning_rate": 0.001,
        "batch_size": 32
    },
    "data": [1, 2, 3],
    "name": "model_v1"
}

new_item = {
    "config": {
        "learning_rate": 0.0005,
        "epochs": 100
    },
    "data": [4, 5],
    "version": 2
}

ARGS = {
    "base_item": base_item,
    "new_item": new_item
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
parse_timestamp Ran
ran, honoured the contract we drafted
thudm/visualagentbench/src/analysis.py
code served (permissive licence) · get_code("99bba02533b726bc")
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 float, 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 datetime
import re
import time

def parse_timestamp(time_str: str) -> float:
    try:
        return float(time_str)
    except:
        pass
    try:
        return datetime.datetime.strptime(time_str, '%Y-%m-%d %H:%M:%S').timestamp()
    except:
        pass
    try:
        return datetime.datetime.strptime(time_str, '%Y-%m-%d').timestamp()
    except:
        pass
    try:
        return datetime.datetime.strptime(time_str, '%Y-%m').timestamp()
    except:
        pass
    num = float(re.findall('[\\d\\.]+', time_str)[0])
    unit = re.findall('[a-zA-Z]+', time_str)[0]
    if unit == 'd':
        delta = num * 24 * 60 * 60
    elif unit == 'h':
        delta = num * 60 * 60
    elif unit == 'm':
        delta = num * 60
    elif unit == 's':
        delta = num
    else:
        raise Exception('Unknown time unit')
    return time.time() - delta

ARGS = {
    'time_str': '5.5d'
}
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 float, 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
get_result Not yet run
unverified (harvested, no recorded run)
thudm/visualagentbench/VAB-WebArena-Lite/new/score.py
code served (permissive licence) · get_code("5372a3148cae304b")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 2 later) and 0 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
deep_merge 2 other papers
1 dated earlier, 1 later
2308.03688 (2023-08)
AgentBench: Evaluating LLMs as Agents
2501.14654 (2025-01)
MedAgentBench: A Realistic Virtual EHR Environment to Benchmark Medical LLM Agents
analyze_output 1 other paper
1 dated earlier
2308.03688 (2023-08)
AgentBench: Evaluating LLMs as Agents
get_result 1 other paper
1 later
2505.16421 (2025-05)
WEBAGENT-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning
parse_timestamp 1 other paper
1 dated earlier
2308.03688 (2023-08)
AgentBench: Evaluating LLMs as Agents

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 Multimodal Models (LMMs) have ushered in a new era in artificial intelligence, merging capabilities in both language and vision to form highly capable Visual Foundation Agents. These agents are postulated to excel across a myriad of tasks, potentially approaching general artificial intelligence. However, existing benchmarks fail to sufficiently challenge or showcase the full potential of LMMs in complex, real-world environments. To address this gap, we introduce VisualAgentBench (VAB), a comprehensive and pioneering benchmark specifically designed to train and evaluate LMMs as visual foundation agents across diverse scenarios, including Embodied, Graphical User Interface, and Visual Design, with tasks formulated to probe the depth of LMMs' understanding and interaction capabilities. Through rigorous testing across nine proprietary LMM APIs and eight open models, we demonstrate the considerable yet still developing agent capabilities of these models. Additionally, VAB constructs a trajectory training set constructed through hybrid methods including Program-based Solvers, LMM Agent Bootstrapping, and Human Demonstrations, promoting substantial performance improvements in LMMs through behavior cloning. Our work not only aims to benchmark existing models but also provides a solid foundation for future development into visual foundation agents. Code, train \& test data, and part of fine-tuned open LMMs are available at \url{https://github.com/THUDM/VisualAgentBench}.

For agents

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

get_harvested_code_for_paper("2408.06327")
get_code_for_paper("2408.06327")
have("2408.06327")

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

curl https://syntology.ai/api/ran/2408.06327.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.06327.svg)](https://syntology.ai/paper/2408.06327)

Connect an agent — have() is free.