SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.03302 · 2023

MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation

arXiv · PDF · Open in the Atlas

Code that ran

5 of 10 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; 0 honoured the contract we drafted, 2 violated it and 1 has no contract verdict; 2 failures of our instrument; 5 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/2310.03302.json · badge: /api/ran/2310.03302.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
snap-stanford/mlagentbench canonical (classifier label)
classifier basis: census
5 ran · 5 no recorded run
FunctionStatusWhere it lives
check_file_read_only Ran
ran; our contract draft was wrong, not the code
snap-stanford/mlagentbench/MLAgentBench/low_level_actions.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("8fbba13275be664e")
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 du-nlp-lab/mlr-copilot, 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.

mkl_error Ran
ran, violated the contract we drafted
snap-stanford/mlagentbench/MLAgentBench/plot.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("53250bc9bc9a4820")
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 du-nlp-lab/mlr-copilot, 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.

oom_error Ran
ran, violated the contract we drafted
snap-stanford/mlagentbench/MLAgentBench/plot.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("357dc7ee438a7576")
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 du-nlp-lab/mlr-copilot, 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.

reflection Ran
ran; our contract draft was wrong, not the code
snap-stanford/mlagentbench/MLAgentBench/high_level_actions.py
code served (permissive licence) · get_code("4f82a282dda31635")
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 os
import tempfile

# Mock the read_file function
def read_file(filename, work_dir='.', **kwargs):
    return "Previous research findings: Model A achieved 85% accuracy. Model B needs tuning."

# Mock the complete_text_fast function
def complete_text_fast(prompt, log_file=None):
    return "Based on the research log, Model A shows promise but needs validation on test set. Model B requires hyperparameter optimization."

# Inject mocks into the module namespace
import sys
current_module = sys.modules[__name__]
current_module.read_file = read_file
current_module.complete_text_fast = complete_text_fast

# Create a temporary work directory
work_dir = tempfile.mkdtemp()

# Create a mock research_log.log file
log_file_path = os.path.join(work_dir, 'research_log.log')
with open(log_file_path, 'w') as f:
    f.write("Previous research findings: Model A achieved 85% accuracy. Model B needs tuning.")

# Create a mock log file for kwargs
mock_log_file = os.path.join(work_dir, 'mock.log')
with open(mock_log_file, 'w') as f:
    f.write("Mock log")

ARGS = {
    'things_to_reflect_on': 'Should we pursue Model A further or invest in Model B improvements?',
    'work_dir': work_dir,
    'research_problem': 'Develop a machine learning model for image classification that achieves >90% accuracy',
    'log_file': mock_log_file
}
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
understand_file Ran
ran on a synthesized input
snap-stanford/mlagentbench/MLAgentBench/high_level_actions.py
code served (permissive licence) · get_code("e78c16077ff8eccc")
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
from unittest.mock import Mock, patch, MagicMock
import tempfile
import os

# Mock the read_file function
def mock_read_file(file_name, work_dir='.', **kwargs):
    return """def hello():
    print('line 1')
    print('line 2')
    print('line 3')
def world():
    print('line 4')
    print('line 5')
    print('line 6')""" + "\n" * 100

# Mock the complete_text_fast function
def mock_complete_text_fast(prompt, log_file=None):
    return f"Mock completion for segment at {log_file}"

# Set up patches
def setup_args():
    with patch('__main__.read_file', side_effect=mock_read_file):
        with patch('__main__.complete_text_fast', side_effect=mock_complete_text_fast):
            args_dict = {
                'file_name': 'test_script.py',
                'things_to_look_for': 'Find all function definitions and their purposes',
                'work_dir': '.',
                'log_file': 'test_log'
            }
            return args_dict

# Build the arguments
ARGS = {
    'file_name': 'example.py',
    'things_to_look_for': 'Find function definitions, imports, and class declarations with their line numbers',
    'work_dir': '.',
    'log_file': 'analysis_log'
}
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
create_benchmark_folder_name Not yet run
unverified (harvested, no recorded run)
snap-stanford/mlagentbench/MLAgentBench/environment.py
code served (permissive licence) · get_code("28af11cb458eb742")
How it ran

no invocation recorded.

edit_script Not yet run
unverified (harvested, no recorded run)
snap-stanford/mlagentbench/MLAgentBench/high_level_actions.py
code served (permissive licence) · get_code("a1cb2a52e65ff361")
How it ran

no invocation recorded.

estimate_tokens Not yet run
unverified (harvested, no recorded run)
snap-stanford/mlagentbench/MLAgentBench/plot.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("9c55f0596ab5430b")
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 du-nlp-lab/mlr-copilot, 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. No invocation recorded.

get_task_info Not yet run
unverified (harvested, no recorded run)
snap-stanford/mlagentbench/MLAgentBench/prepare_task.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("2868580e4a832fe7")
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 du-nlp-lab/mlr-copilot, 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. No invocation recorded.

normalize_args_kwargs Not yet run
unverified (harvested, no recorded run)
snap-stanford/mlagentbench/MLAgentBench/low_level_actions.py
pointer only (licence: MIT); the stored copy get_code serves, from du-nlp-lab/mlr-copilot, is pointer only (licence: NONE) · get_code("d6d0c6d7bde0bdcc")
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 du-nlp-lab/mlr-copilot, 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. No invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 6 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 6 later) and 4 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
check_file_read_only 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
estimate_tokens 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
get_task_info 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
mkl_error 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
normalize_args_kwargs 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents
oom_error 1 other paper
1 later
2408.14033 (2024-08)
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models 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

A central aspect of machine learning research is experimentation, the process of designing and running experiments, analyzing the results, and iterating towards some positive outcome (e.g., improving accuracy). Could agents driven by powerful language models perform machine learning experimentation effectively? To answer this question, we introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We then construct an agent that can perform ML experimentation based on ReAct framework. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate. It can build compelling ML models over many tasks in MLAgentBench with 37.5% average success rate. Our agents also display highly interpretable plans and actions. However, the success rates vary considerably; they span from 100% on well-established older datasets to as low as 0% on recent Kaggle challenges created potentially after the underlying LM was trained. Finally, we identify several key challenges for LM-based agents such as long-term planning and reducing hallucination. Our code is released at https://github.com/snap-stanford/MLAgentBench.

For agents

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

get_harvested_code_for_paper("2310.03302")
get_code_for_paper("2310.03302")
have("2310.03302")

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

curl https://syntology.ai/api/ran/2310.03302.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.03302.svg)](https://syntology.ai/paper/2310.03302)

Connect an agent — have() is free.