SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.17168 · ACL · 2024

Benchmarking Data Science Agents

Yuge Zhang, Kan Ren, Yuqing Yang, Nan Chen, Qiyang Jiang, Xingyu Han, Jupyter Ai

arXiv · PDF · Open in the Atlas

Code that ran

4 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 3 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/2402.17168.json · badge: /api/ran/2402.17168.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
metacopilot/dseval canonical (classifier label)
classifier basis: census
4 ran · 3 no recorded run
FunctionStatusWhere it lives
add_handler Ran
ran; our contract draft was wrong, not the code
metacopilot/dseval/dseval/logging.py
code served (permissive licence) · get_code("80aaeed8b41d1da8")
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 StreamHandler. 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 logging
import sys
from pathlib import Path
from typing import Optional

# Initialize the global _file_handlers dict that the function uses
_file_handlers = {}

# Create a logger for testing
test_logger = logging.getLogger("test_logger")

# Define ARGS with all parameters of add_handler
ARGS = {
    "logger": test_logger,
    "file": None,
}
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 StreamHandler — a StreamHandler output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a StreamHandler 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
add_indent Ran
ran; our contract draft was wrong, not the code
metacopilot/dseval/dseval/utils.py
code served (permissive licence) · get_code("eb53b42301057566")
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

# Simple test cases for add_indent function
text_simple = "hello\nworld"
text_multiline = "line1\nline2\nline3"
text_single = "single line"
text_empty = ""

ARGS = {
    "text": "hello\nworld\nfoo",
    "indent": 2
}
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
available_results Ran
ran; our contract draft was wrong, not the code
metacopilot/dseval/dseval/browser.py
code served (permissive licence) · get_code("995319536acf66b7")
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 tempfile
from pathlib import Path
import json

# Create a temporary directory structure with .jsonl files
temp_dir = tempfile.TemporaryDirectory()
root_dir = Path(temp_dir.name)

# Create some subdirectories and .jsonl files
(root_dir / "subdir1").mkdir(exist_ok=True)
(root_dir / "subdir2").mkdir(exist_ok=True)
(root_dir / "subdir1" / "subdir3").mkdir(exist_ok=True)

# Create some .jsonl files
(root_dir / "results1.jsonl").write_text('{"key": "value"}\n')
(root_dir / "subdir1" / "results2.jsonl").write_text('{"key": "value"}\n')
(root_dir / "subdir1" / "subdir3" / "results3.jsonl").write_text('{"key": "value"}\n')
(root_dir / "subdir2" / "results4.jsonl").write_text('{"key": "value"}\n')

ARGS = {
    "root_dir": root_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
categorize_comparison_failure Ran
ran; our contract draft was wrong, not the code
metacopilot/dseval/dseval/validator.py
code served (permissive licence) · get_code("2b525d76f753e7a8")
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 Subverdict. 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 torch.nn as nn

ARGS = {
    "failure": "shape mismatch: expected (2, 3) but got (2, 4)"
}
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 Subverdict — a Subverdict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Subverdict 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
exec_code Not yet run
unverified (harvested, no recorded run)
metacopilot/dseval/dseval/utils.py
code served (permissive licence) · get_code("7af9f9ee08a52c95")
How it ran

no invocation recorded.

get_temporary_file_name Not yet run
unverified (harvested, no recorded run)
metacopilot/dseval/dseval/utils.py
code served (permissive licence) · get_code("6a28b96a048d0343")
How it ran

no invocation recorded.

validator_comments_to_verdict Not yet run
unverified (harvested, no recorded run)
metacopilot/dseval/dseval/validator.py
code served (permissive licence) · get_code("399a475723bfd916")
How it ran

no invocation recorded.

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, 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 7 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

In the era of data-driven decision-making, the complexity of data analysis necessitates advanced expertise and tools of data science, presenting significant challenges even for specialists. Large Language Models (LLMs) have emerged as promising aids as data science agents, assisting humans in data analysis and processing. Yet their practical efficacy remains constrained by the varied demands of realworld applications and complicated analytical process. In this paper, we introduce DSEvala novel evaluation paradigm, as well as a series of innovative benchmarks tailored for assessing the performance of these agents throughout the entire data science lifecycle. Incorporating a novel bootstrapped annotation method, we streamline dataset preparation, improve the evaluation coverage, and expand benchmarking comprehensiveness. Our findings uncover prevalent obstacles and provide critical insights to inform future advancements in the field. * * Correspondence to Kan Ren. * Source code and data are available at https://github. com/MetaCopilot/dseval. In-memory Data country landArea pop2010 pop2023 pop2050 Calculate the population density for each country in 2023 and 2050. Result should be a new frame with "Country" as the index and "2023 Density" and "2050 Density" as the columns.

For agents

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

get_harvested_code_for_paper("2402.17168")
get_code_for_paper("2402.17168")
have("2402.17168")

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

curl https://syntology.ai/api/ran/2402.17168.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.17168.svg)](https://syntology.ai/paper/2402.17168)

Connect an agent — have() is free.