SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.11470 · 2024

Beyond Correctness: Benchmarking Multi-dimensional Code Generation for Large Language Models

arXiv · PDF · Open in the Atlas

Code that ran

12 of 16 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; 10 failures of our instrument; 4 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/2407.11470.json · badge: /api/ran/2407.11470.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
jszheng21/race canonical (classifier label)
classifier basis: census
12 ran · 4 no recorded run
FunctionStatusWhere it lives
capitalize_after_underscore Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/evaluate_pipeline_evalplus.py
code served (permissive licence) · get_code("52e3021e90c7e1f4")
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 re

def capitalize_after_underscore(s):
    return re.sub('_(\\w)', lambda x: x.group(1).upper(), s)

ARGS = {
    's': 'hello_world_test_string'
}
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
compatible_eval_result Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/evaluation_utils_evalplus.py
code served (permissive licence) · get_code("0fe1c97ce512eef9")
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 torch

# Construct a valid argument for the compatible_eval_result function
# The function expects a dict with an 'eval' key containing task results
# It processes 'files' keys by converting them to 'nfiles' (count of files)

results = {
    'eval': {
        'task1': {
            'files': ['file1.txt', 'file2.txt', 'file3.txt'],
            'accuracy': 0.95
        },
        'task2': {
            'nfiles': 5,
            'loss': 0.1
        },
        'task3': {
            'files': ['a.txt'],
            'score': 0.87
        }
    }
}

ARGS = {
    'results': results
}
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 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
construct_prompt_customized Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codegen/prompts.py
code served (permissive licence) · get_code("f77f0258d19b7231")
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
from transformers import AutoTokenizer

# Create a minimal tokenizer stand-in class that has the required attributes/methods
class MinimalTokenizer:
    def __init__(self, has_chat_template=False):
        self.chat_template = "{% for message in messages %}{{ message['role'] }}: {{ message['content'] }}\n{% endfor %}" if has_chat_template else None
    
    def apply_chat_template(self, messages, tokenize=False):
        # Simple implementation that formats messages
        result = ""
        for message in messages:
            result += f"{message['role']}: {message['content']}\n"
        return result

# Test case 1: tokenizer without chat_template
tokenizer_no_chat = MinimalTokenizer(has_chat_template=False)

# Test case 2: tokenizer with chat_template
tokenizer_with_chat = MinimalTokenizer(has_chat_template=True)

# Create test arguments
ARGS = {
    'prompt': 'Write a function that adds two numbers and returns the result.',
    'dim': 'correctness',
    'tokenizer': tokenizer_no_chat
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — 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
estimate_pass_at_k Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/evaluation_utils_evalplus.py
code served (permissive licence) · get_code("5a283b50e4b14e37")
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 ndarray, ndim 1, dtype family float, shape (3,) 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 numpy as np

# The function estimate_pass_at_k takes:
# - num_samples: Union[int, List[int], np.ndarray] - number of samples per problem
# - num_correct: Union[List[int], np.ndarray] - number of correct samples per problem
# - k: int - the k value for pass@k

# Let's construct valid arguments
num_samples_val = 10  # Can be int, list, or np.ndarray
num_correct_val = np.array([3, 5, 7])  # Array of correct counts for 3 problems
k_val = 2  # The k parameter for pass@k

ARGS = {
    "num_samples": num_samples_val,
    "num_correct": num_correct_val,
    "k": k_val
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (3,) — shape (3,) 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
extra_eos_for_direct_completion Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codegen/model.py
code served (permissive licence) · get_code("cff3266598686087")
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 torch
import numpy as np
from typing import List

# The function takes a single parameter: dataset
# It expects a string that will be checked with .lower()
# and compared against specific dataset names

ARGS = {
    'dataset': 'humaneval'
}
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 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
extract_class_names Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/metrics_utils.py
code served (permissive licence) · get_code("9f63a082cf9ac9d7")
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 ast

def extract_class_names(code):
    try:
        code_bytes = bytes(code, 'utf8')
        tree = ast.parse(code_bytes)
    except SyntaxError:
        return []
    class_names = [node.name for node in ast.walk(tree) if isinstance(node, ast.ClassDef)]
    return class_names

# Setup code to construct valid arguments
code_sample = """
class MyFirstClass:
    def __init__(self):
        self.value = 42

class MySecondClass:
    def method(self):
        return "hello"

def some_function():
    pass
"""

ARGS = {
    "code": code_sample
}
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
extract_function_names Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/metrics_utils.py
code served (permissive licence) · get_code("c496a7988b94902a")
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 ast

def extract_function_names(code):
    try:
        code_bytes = bytes(code, 'utf8')
        tree = ast.parse(code_bytes)
    except SyntaxError:
        return []
    names = []
    for node in ast.walk(tree):
        if isinstance(node, ast.FunctionDef):
            names.append(node.name)
    return names

# Setup code
code_sample = """
def hello():
    pass

def world():
    return 42

def foo_bar():
    x = 1
"""

ARGS = {
    "code": code_sample
}
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
extract_variable_names Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codeeval/metrics_utils.py
code served (permissive licence) · get_code("93545955a0c6e69e")
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 set. 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 ast

code_sample_1 = """
x = 5
y = 10
z = x + y
"""

code_sample_2 = """
def my_function(a, b, c):
    result = a + b
    return result
"""

code_sample_3 = """
for i, j in [(1, 2), (3, 4)]:
    print(i, j)
"""

code_sample_4 = """
x, y = 1, 2
"""

code_sample_5 = """
invalid syntax here ][
"""

ARGS = {
    "code": code_sample_1
}
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 set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set 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_leading_spaces Ran
ran, honoured the contract we drafted
jszheng21/race/race/codegen/utils.py
code served (permissive licence) · get_code("1f9ac358e6dcdc7d")
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.

import torch
import numpy as np

def get_leading_spaces(string):
    return len(string) - len(string.lstrip())

ARGS = {
    "string": "    hello world"
}
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 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
is_floats Ran
ran, violated the contract we drafted
jszheng21/race/race/codeeval/evaluation_utils_evalplus.py
code served (permissive licence) · get_code("0b9209ad7ddba928")
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 numpy as np

# The function is_floats takes one parameter: x
# It checks if x is a float, a list/tuple of floats, or a numpy array of float dtype
# We need to provide valid test cases for this function

ARGS = {
    "x": 3.14
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
retry_get_openai_response Ran
ran; our contract draft was wrong, not the code
jszheng21/race/race/codegen/utils.py
code served (permissive licence) · get_code("4ed59fa1ec6ea391")
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 Mock. 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 openai
from unittest.mock import Mock, MagicMock

# Create a mock OpenAI client
mock_client = Mock(spec=openai.Client)

# Create a mock ChatCompletion response
mock_choice = Mock()
mock_choice.text = "This is a test response"

mock_completion = Mock(spec=openai.types.chat.ChatCompletion)
mock_completion.choices = [mock_choice]

# Configure the client to return our mock completion
mock_client.chat.completions.create.return_value = mock_completion

ARGS = {
    'client': mock_client,
    'model': 'gpt-3.5-turbo',
    'user_prompt': 'What is the capital of France?',
    'system_prompt': 'You are a helpful assistant.',
    'max_tokens': 100,
    'temperature': 0.7,
    'n': 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
openai, tenacity — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Mock — a Mock output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Mock 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
trusted_exec Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
jszheng21/race/race/codeeval/execution_utils_evalplus.py
code served (permissive licence) · get_code("9e1a929f64001008")
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 int; 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 torch
import time
from copy import deepcopy

torch.manual_seed(0)

# Create a simple function to be executed
sample_code = """
def add_numbers(a, b):
    return a + b

def check_value(x):
    return x is not None
"""

# Create inputs - list of tuples/lists of arguments for each function call
inputs = [
    (2, 3),
    (5, 7),
    (10, 15)
]

# Entry point is the function name to call
entry_point = "add_numbers"

# Record time flag
record_time = False

# Output not none flag
output_not_none = False

ARGS = {
    "code": sample_code,
    "inputs": inputs,
    "entry_point": entry_point,
    "record_time": record_time,
    "output_not_none": output_not_none
}
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 list, ndim 1, dtype family int — 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
construct_prompt_evalplus Not yet run
unverified (harvested, no recorded run)
jszheng21/race/race/codegen/prompts.py
code served (permissive licence) · get_code("7d55790d6fc57e9d")
How it ran

no invocation recorded.

construct_prompt_leetcode_efficiency Not yet run
unverified (harvested, no recorded run)
jszheng21/race/race/codegen/prompts.py
code served (permissive licence) · get_code("5b9243b85fe4c3ad")
How it ran

no invocation recorded.

retry_get_genai_response Not yet run
unverified (harvested, no recorded run)
jszheng21/race/race/codegen/utils.py
code served (permissive licence) · get_code("7ae0c34bc3033b6d")
How it ran

no invocation recorded.

test_time_and_memory_usage Not yet run
unverified (harvested, no recorded run)
jszheng21/race/race/codeeval/execution_utils.py
code served (permissive licence) · get_code("63e64eea84bc55fa")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 12 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 12 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
estimate_pass_at_k 20 other papers
14 dated earlier, 1 the same month, 5 later
2107.03374 (2021-07)
Evaluating Large Language Models Trained on Code
2207.14255 (2022-07)
Efficient Training of Language Models to Fill in the Middle
2210.14868 (2022-10)
Multi-lingual Evaluation of Code Generation Models
and 17 more not named here
capitalize_after_underscore 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
compatible_eval_result 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
construct_prompt_customized 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
construct_prompt_evalplus 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
construct_prompt_leetcode_efficiency 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
extra_eos_for_direct_completion 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
extract_class_names 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
extract_function_names 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
extract_variable_names 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
is_floats 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling
trusted_exec 1 other paper
1 later
2411.17501 (2024-11)
The Limits of Inference Scaling Through Resampling

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 recent years, researchers have proposed numerous benchmarks to evaluate the impressive coding capabilities of large language models (LLMs). However, current benchmarks primarily assess the accuracy of LLM-generated code, while neglecting other critical dimensions that also significantly impact code quality in real-world development. Moreover, relying exclusively on correctness as the guiding metric renders LLMs susceptible to data contamination. Therefore, this paper proposes the RACE benchmark, which comprehensively evaluates the quality of code generated by LLMs across 4 dimensions: Readability, mAintainability, Correctness, and Efficiency. Specifically, considering the demand-dependent nature of dimensions beyond correctness, we design various types of user requirements for each dimension to assess the model's ability to generate correct code that also meets user demands. We analyze 28 representative LLMs based on RACE and find that: 1) current correctness-centric benchmarks fail to capture the multifaceted requirements of code in real-world scenarios, while RACE provides a comprehensive evaluation that reveals the defects of LLMs across multiple dimensions; 2) the RACE benchmark serves as an effective tool for resisting the risk of data contamination; 3) even the most advanced code LLMs still encounter significant challenges in customized requirements involving complex instructions; 4) most LLMs exhibit an inherent preference for specific coding style. These findings highlight the need for a multidimensional evaluation of code LLMs, emphasizing metrics beyond correctness for real-world applications. Future efforts should aim to develop novel learning algorithms to enhance code generation under varied constraints and improve coverage and usability for diverse user needs.

For agents

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

get_harvested_code_for_paper("2407.11470")
get_code_for_paper("2407.11470")
have("2407.11470")

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

curl https://syntology.ai/api/ran/2407.11470.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.11470.svg)](https://syntology.ai/paper/2407.11470)

Connect an agent — have() is free.