SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.20550 · 2024

LLM Hallucinations in Practical Code Generation: Phenomena, Mechanism, and Mitigation

arXiv · PDF · Open in the Atlas

Code that ran

5 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, 1 violated it and 0 have no contract verdict; 4 failures of our instrument; 2 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/2409.20550.json · badge: /api/ran/2409.20550.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
deepsoftwareanalytics/llmcodinghallucination canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
clean_code Ran
ran; our contract draft was wrong, not the code
deepsoftwareanalytics/llmcodinghallucination/eval_original.py
code served (permissive licence) · get_code("c8230eeeb2532352")
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 clean_code(code: str):
    lines = code.split('\n')
    lines = [line for line in lines if not line.strip().startswith('#')]
    code = '\n'.join(lines)
    return re.sub("'''(.*)'''", '', code)

# Setup code to construct valid arguments
sample_code = """# This is a comment
def hello():
    '''This is a docstring'''
    print('Hello')
# Another comment
x = 5
'''Another docstring'''
y = 10
"""

ARGS = {
    "code": sample_code
}
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
clean_docstring Ran
ran; our contract draft was wrong, not the code
deepsoftwareanalytics/llmcodinghallucination/eval_original.py
code served (permissive licence) · get_code("c4a538a65aa5b254")
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 clean_docstring(docstring):
    docstring = docstring.split('\n\n')[0].strip()
    docstring = docstring.split(':param')[0].strip()
    docstring = docstring.split(':return')[0].strip()
    docstring = re.sub('\\s*\\n\\s*[A-Z].*$', '', docstring, re.S)
    docstring = re.sub('\\s+', ' ', docstring)
    docstring = docstring.split('. ')[0]
    return docstring

ARGS = {
    'docstring': """This is a sample docstring for testing.
    
    It has multiple paragraphs.
    
    :param arg1: Description of arg1
    :param arg2: Description of arg2
    :return: Something useful"""
}
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
clean_pad Ran
ran; our contract draft was wrong, not the code
deepsoftwareanalytics/llmcodinghallucination/model.py
code served (permissive licence) · get_code("1a8cc6f72a5c0992")
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 clean_pad(code: str):
    code = re.sub(' ?%s ?' % re.escape('<pad>'), '', code)
    return code

# The function takes a single string parameter and removes <pad> tokens
# with optional surrounding spaces using regex substitution

ARGS = {
    "code": "def hello ( ) : <pad> return <pad> 42"
}
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
clean_str Ran
ran; our contract draft was wrong, not the code
deepsoftwareanalytics/llmcodinghallucination/model.py
code served (permissive licence) · get_code("c7261b10bd59d790")
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 clean_str(code):
    code = re.sub("'(.*?)'", '', code)
    code = re.sub('"(.*?)"', '', code)
    return code.strip()

# The function takes a single parameter 'code' which is a string
# It performs regex substitutions to remove single and double quoted strings
# Then strips whitespace and returns the result

ARGS = {
    "code": "print('hello') and print(\"world\") and x = 5"
}
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
compute_EM Ran
ran, violated the contract we drafted
deepsoftwareanalytics/llmcodinghallucination/compute_score.py
code served (permissive licence) · get_code("fae4a6b218aa55fb")
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)

# target: a multi-line string representing the target code/output
target = """def hello():
    print('world')
    return 42"""

# predictions: a list of strings, each representing a predicted code/output
predictions = [
    """def hello():
    print('world')
    return 42""",
    """def hello():
    print('goodbye')
    return 0""",
    """def foo():
    return 1"""
]

# passk: an integer specifying how many predictions to check
passk = 2

ARGS = {
    "target": target,
    "predictions": predictions,
    "passk": passk
}
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
init_codegen Not yet run
unverified (harvested, no recorded run)
deepsoftwareanalytics/llmcodinghallucination/model.py
code served (permissive licence) · get_code("1673a2948bb1d361")
How it ran

no invocation recorded.

predict Not yet run
unverified (harvested, no recorded run)
deepsoftwareanalytics/llmcodinghallucination/eval_original.py
code served (permissive licence) · get_code("297de4419cd21f7b")
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

Code generation aims to automatically generate code from input requirements, significantly enhancing development efficiency. Recent large language models (LLMs) based approaches have shown promising results and revolutionized code generation task. Despite the promising performance, LLMs often generate contents with hallucinations, especially for the code generation scenario requiring the handling of complex contextual dependencies in practical development process. Although previous study has analyzed hallucinations in LLM-powered code generation, the study is limited to standalone function generation. In this paper, we conduct an empirical study to study the phenomena, mechanism, and mitigation of LLM hallucinations within more practical and complex development contexts in repository-level generation scenario. First, we manually examine the code generation results from six mainstream LLMs to establish a hallucination taxonomy of LLM-generated code. Next, we elaborate on the phenomenon of hallucinations, analyze their distribution across different models. We then analyze causes of hallucinations and identify four potential factors contributing to hallucinations. Finally, we propose an RAG-based mitigation method, which demonstrates consistent effectiveness in all studied LLMs. The replication package including code, data, and experimental results is available at https://github.com/DeepSoftwareAnalytics/LLMCodingHallucination

For agents

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

get_harvested_code_for_paper("2409.20550")
get_code_for_paper("2409.20550")
have("2409.20550")

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

curl https://syntology.ai/api/ran/2409.20550.json

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

[![Syntology run record](https://syntology.ai/api/ran/2409.20550.svg)](https://syntology.ai/paper/2409.20550)

Connect an agent — have() is free.