SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2401.12554 · 2024

Can Large Language Models Write Parallel Code?

arXiv · PDF · Open in the Atlas

Code that ran

18 of 19 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; 6 honoured the contract we drafted, 5 violated it and 0 have no contract verdict; 7 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/2401.12554.json · badge: /api/ran/2401.12554.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
parallelcodefoundry/ParEval canonical (classifier label)
classifier basis: census
18 ran · 1 no recorded run
FunctionStatusWhere it lives
all_equal Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/drivers/util.py
code served (permissive licence) · get_code("de2d759b067567af")
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)

# The function all_equal takes an iterable and checks if all values are equal
# by converting to a set and checking if the set has <= 1 element

# Test case 1: All equal elements
iterable_all_equal = [1, 1, 1, 1]

ARGS = {
    "iterable": iterable_all_equal
}
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
already_has_results Ran
ran, violated the contract we drafted
parallelcodefoundry/ParEval/drivers/run-all.py
code served (permissive licence) · get_code("e7bbb8f061c2443c")
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

# Construct a prompt dict with results stored as dicts
ARGS = {
    'prompt': {
        'name': 'test_prompt',
        'outputs': [
            {'result': 'output1', 'score': 0.95},
            {'result': 'output2', 'score': 0.87}
        ]
    }
}
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
any_in Ran
ran, violated the contract we drafted
parallelcodefoundry/ParEval/analysis/bin-the-stack.py
code served (permissive licence) · get_code("0938e73682f13d8b")
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
import numpy as np
from typing import List

torch.manual_seed(0)

# The function any_in takes:
# - s: str - a string to search in
# - substrs: List[str] - a list of substrings to search for

s = "hello world"
substrs = ["world", "foo", "bar"]

ARGS = {
    "s": s,
    "substrs": substrs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy, 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
buildk Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/analysis/metrics.py
code served (permissive licence) · get_code("98ceec1f11b8e7a0")
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 DataFrame. 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 pandas as pd

# Create a sample DataFrame with the required columns
data = {
    'name': ['model_a', 'model_a', 'model_b', 'model_b', 'model_c', 'model_c'],
    'parallelism_model': ['data', 'data', 'model', 'model', 'data', 'model'],
    'problem_type': ['classification', 'regression', 'classification', 'regression', 'classification', 'regression'],
    'did_build': [1, 0, 1, 1, 0, 1]
}

df = pd.DataFrame(data)

k = 2

ARGS = {
    'df': df,
    'k': k
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
clean_output Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/generate/clean-outputs.py
code served (permissive licence) · get_code("b588819cf6250c27")
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

# Define ARGS for clean_output function
ARGS = {
    "prompt": "def hello():\n    print('world')",
    "output": "print('hello')"
}
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
efficiencyk Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/analysis/metrics-scaling.py
code served (permissive licence) · get_code("8d9bdc243671a23f")
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 DataFrame. 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 pandas as pd

# Create a sample dataframe with the required columns
np.random.seed(0)

# Create base data
n_rows = 12
df = pd.DataFrame({
    'is_valid': [True] * n_rows,
    'n': [2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3],
    'name': ['task_a'] * 6 + ['task_b'] * 6,
    'parallelism_model': ['model_x'] * 3 + ['model_y'] * 3 + ['model_x'] * 3 + ['model_y'] * 3,
    'problem_type': ['type_1'] * 3 + ['type_2'] * 3 + ['type_1'] * 3 + ['type_2'] * 3,
    'output_idx': list(range(3)) * 4,
    'runtime': np.random.uniform(0.1, 2.0, n_rows),
    'best_sequential_runtime': np.random.uniform(0.5, 3.0, n_rows),
})

ARGS = {
    'df': df,
    'k': 2,
    'n': 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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_correctness_df Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/analysis/metrics.py
code served (permissive licence) · get_code("e011cbcff09dc44b")
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 DataFrame. 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 pandas as pd

# Create a sample DataFrame with the required columns
df = pd.DataFrame({
    'name': ['model_a', 'model_a', 'model_a', 'model_b', 'model_b', 'model_b'],
    'parallelism_model': ['dp', 'dp', 'tp', 'dp', 'tp', 'tp'],
    'output_idx': [0, 0, 1, 0, 0, 1],
    'is_valid': [True, True, False, True, True, True],
    'problem_type': ['classification', 'classification', 'classification', 'regression', 'regression', 'regression']
})

ARGS = {
    'df': df
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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_env_var Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/generate/generate-gemini.py
code served (permissive licence) · get_code("2f1ce38ddef63dc9")
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 os

# Set an environment variable for the function to retrieve
os.environ['TEST_VAR'] = 'test_value'

# Define arguments for get_env_var
ARGS = {
    'name': 'TEST_VAR'
}
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
get_function_name Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/generate/generate-gemini.py
code served (permissive licence) · get_code("fc1387d6dfa3f110")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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

GPU_FUNCTION_NAME_PATTERN = re.compile('__global__ void ([a-zA-Z0-9_]+)\\(')
CPU_FUNCTION_NAME_PATTERN = re.compile('\\s*[a-zA-Z_]+ ([a-zA-Z0-9_]+)\\(')

def get_function_name(prompt: str, execution_model: str) -> str:
    if execution_model in ['cuda', 'hip']:
        match = GPU_FUNCTION_NAME_PATTERN.match(prompt.splitlines()[-1])
    else:
        match = CPU_FUNCTION_NAME_PATTERN.match(prompt.splitlines()[-1])
    if match is None:
        raise ValueError(f'Could not find function name in prompt: {prompt}')
    return match.group(1)


ARGS = {
    'prompt': '__global__ void kernel_function(',
    'execution_model': 'cuda'
}
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
get_max_tokens_per_second Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/generate/generate-gemini.py
code served (permissive licence) · get_code("60484b445abb73b5")
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 torch
import torch.nn as nn
from typing import Optional

# Set seed for reproducibility
torch.manual_seed(0)

# The function takes a model parameter (string) and checks if it equals 'gemini-1.0-pro'
# It returns tokens per second or None
# The model parameter is just a string, so we pass a valid model name

ARGS = {
    'model': 'gemini-1.0-pro'
}
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 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_max_tokens_per_second Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/generate/generate-openai.py
code served (permissive licence) · get_code("85d8ccfa5ba4fb7b")
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 torch
from typing import Optional

def get_max_tokens_per_second(model: str) -> Optional[int]:
    """ rates limites as of January 2024 """
    if model == 'gpt-3.5-turbo-1106':
        tokens_per_minute = 60000
        return tokens_per_minute / 60
    elif model == 'gpt-4-1106-preview':
        tokens_per_minute = 150000
        return tokens_per_minute / 60
    else:
        return None

ARGS = {
    "model": "gpt-3.5-turbo-1106"
}
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 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
has_balanced_brackets Ran
ran, violated the contract we drafted
parallelcodefoundry/ParEval/generate/clean-outputs.py
code served (permissive licence) · get_code("d55502b35fc3f4c1")
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.

def has_balanced_brackets(s: str) -> bool:
    stack = []
    for c in s:
        if c == '{':
            stack.append(c)
        elif c == '}':
            if len(stack) == 0:
                return False
            stack.pop()
    return len(stack) == 0


ARGS = {
    's': '{{}}{}{{{}}}}'
}
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 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
has_outputs Ran
ran, violated the contract we drafted
parallelcodefoundry/ParEval/analysis/create-dataframe.py
code served (permissive licence) · get_code("311658ebceb020fe")
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)

# Test case 1: Valid prompt with outputs containing dicts
prompt_valid = {
    'outputs': [
        {'result': 'value1'},
        {'result': 'value2'}
    ]
}

ARGS = {
    'prompt': prompt_valid
}
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
load_json Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/drivers/util.py
code served (permissive licence) · get_code("9430b9d58bca314d")
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.

import json
import tempfile
import os
from os import PathLike

# Create a temporary JSON file with some test data
temp_dir = tempfile.mkdtemp()
test_json_path = os.path.join(temp_dir, "test.json")

test_data = {
    "key1": "value1",
    "key2": 42,
    "key3": ["list", "of", "items"],
    "key4": {"nested": "dict"}
}

with open(test_json_path, 'w') as f:
    json.dump(test_data, f)

ARGS = {
    "fpath": test_json_path
}
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
nCr Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/analysis/metrics-scaling.py
code served (permissive licence) · get_code("c1d07cc2f1b9ecf6")
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 int, ndim 0, dtype family int, 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.

from math import comb

def nCr(n: int, r: int) -> int:
    if n < r:
        return 1
    return comb(n, r)

ARGS = {
    "n": 5,
    "r": 2
}
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 int, ndim 0, dtype family int, 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
read_json Ran
ran; our contract draft was wrong, not the code
parallelcodefoundry/ParEval/analysis/bin-the-stack.py
code served (permissive licence) · get_code("24736464e5b48a38")
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.

import json
import tempfile
import os

# Create a temporary JSON file
temp_dir = tempfile.mkdtemp()
temp_json_path = os.path.join(temp_dir, "test.json")

# Write a valid JSON file
test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "data"}
}

with open(temp_json_path, 'w') as f:
    json.dump(test_data, f)

ARGS = {
    "fpath": temp_json_path
}
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
speedupk Ran
ran, honoured the contract we drafted
parallelcodefoundry/ParEval/analysis/metrics-scaling.py
code served (permissive licence) · get_code("084d0107c87254df")
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 DataFrame. 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 pandas as pd
import numpy as np

# Create a sample DataFrame with the required structure
data = {
    'is_valid': [True, True, True, True, True, True, True, True],
    'n': [10, 10, 10, 10, 20, 20, 20, 20],
    'name': ['task1', 'task1', 'task2', 'task2', 'task1', 'task1', 'task2', 'task2'],
    'parallelism_model': ['model_a', 'model_a', 'model_b', 'model_b', 'model_a', 'model_a', 'model_b', 'model_b'],
    'output_idx': [0, 1, 0, 1, 0, 1, 0, 1],
    'problem_type': ['type_x', 'type_x', 'type_y', 'type_y', 'type_x', 'type_x', 'type_y', 'type_y'],
    'best_sequential_runtime': [1.5, 1.5, 2.0, 2.0, 1.5, 1.5, 2.0, 2.0],
    'runtime': [0.5, 0.6, 0.7, 0.8, 0.5, 0.6, 0.7, 0.8],
}

df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'k': 3,
    'n': 10,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
uses_mpi Ran
ran, violated the contract we drafted
parallelcodefoundry/ParEval/analysis/bin-the-stack.py
code served (permissive licence) · get_code("6eee785f978e2422")
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
import numpy as np
from numba import njit, jit
from typing import List

@njit
def any_in(s: str, substrs: List[str]) -> bool:
    for substr in substrs:
        if substr in s:
            return True
    return False

@njit
def uses_mpi(contents: str, language: str) -> bool:
    if language in ['C', 'C++']:
        return any_in(contents, ['MPI_', '#include <mpi.h>', '#include "mpi.h"'])
    elif language == 'FORTRAN':
        return any_in(contents, ['MPI_', "include 'mpif.h'", 'include "mpif.h"'])
    elif language == 'Python':
        return any_in(contents, ['mpi4py'])
    return False

ARGS = {
    'contents': '#include <mpi.h>\nint main() { MPI_Init(); }',
    'language': 'C'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy, 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
await_input Not yet run
unverified (harvested, no recorded run)
parallelcodefoundry/ParEval/drivers/util.py
code served (permissive licence) · get_code("9c46438f56d92741")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 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 19 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 are increasingly becoming a popular tool for software development. Their ability to model and generate source code has been demonstrated in a variety of contexts, including code completion, summarization, translation, and lookup. However, they often struggle to generate code for complex programs. In this paper, we study the capabilities of state-of-the-art language models to generate parallel code. In order to evaluate language models, we create a benchmark, ParEval, consisting of prompts that represent 420 different coding tasks related to scientific and parallel computing. We use ParEval to evaluate the effectiveness of several state-of-the-art open- and closed-source language models on these tasks. We introduce novel metrics for evaluating the performance of generated code, and use them to explore how well each large language model performs for 12 different computational problem types and six different parallel programming models.

For agents

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

get_harvested_code_for_paper("2401.12554")
get_code_for_paper("2401.12554")
have("2401.12554")

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

curl https://syntology.ai/api/ran/2401.12554.json

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

[![Syntology run record](https://syntology.ai/api/ran/2401.12554.svg)](https://syntology.ai/paper/2401.12554)

Connect an agent — have() is free.