SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.16801 · 2024

RES-Q: Evaluating Code-Editing Large Language Model Systems at the Repository Scale

arXiv · PDF · Open in the Atlas

Code that ran

6 of 6 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; 6 failures of our instrument; 0 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/2406.16801.json · badge: /api/ran/2406.16801.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
qurrent-ai/res-q canonical (classifier label)
classifier basis: census
6 ran · 0 no recorded run
FunctionStatusWhere it lives
extract_modified_files Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/git/utils.py
code served (permissive licence) · get_code("48c807f6feeb0909")
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 re
from typing import List

def extract_modified_files(patch: str) -> List[str]:
    """
    Return a list of files in the diff where more than spaces/newlines are changed
    """
    file_path_regex = re.compile('^diff --git a\\/(.*?) b\\/(.*?)$', re.MULTILINE)
    change_line_regex = re.compile('^[\\+\\-](?!\\+\\+|\\-\\-)(.*)$', re.MULTILINE)
    modified_files = []
    current_file = None
    found_significant_change = False
    for line in patch.splitlines():
        file_path_match = file_path_regex.match(line)
        change_line_match = change_line_regex.match(line)
        if file_path_match:
            if current_file and found_significant_change:
                modified_files.append(current_file)
            current_file = file_path_match.group(1)
            found_significant_change = False
        elif change_line_match:
            if change_line_match.group(1).strip():
                found_significant_change = True
    if current_file and found_significant_change:
        modified_files.append(current_file)
    return modified_files


patch_example = """diff --git a/src/main.py b/src/main.py
index 1234567..abcdefg 100644
--- a/src/main.py
+++ b/src/main.py
@@ -10,6 +10,7 @@ def hello():
     print("hello")
+    print("world")
     return True
diff --git a/src/utils.py b/src/utils.py
index 9876543..fedcba9 100644
--- a/src/utils.py
+++ b/src/utils.py
@@ -5,3 +5,4 @@ def helper():
-    old_line = 123
+    new_line = 456
"""

ARGS = {
    "patch": patch_example
}
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
filter_binary_patch Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/git/utils.py
code served (permissive licence) · get_code("76b18236be6ad0d9")
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 filter_binary_patch(patch: str) -> str:
    """
    Filter out binary patches from the patch
    """
    lines = patch.splitlines(keepends=True)
    filtered_lines = []
    in_binary_section = False
    binary_patch_start = re.compile('^GIT binary patch$')
    binary_patch_header = re.compile('^index ')
    binary_files_differ = re.compile('^Binary files .* and .* differ$')
    git_diff_line = re.compile('^diff --git ')
    i = 0
    while i < len(lines):
        line = lines[i]
        if binary_patch_start.match(line):
            in_binary_section = True
            i += 1
            continue
        elif binary_patch_header.match(line):
            i += 1
            continue
        elif binary_files_differ.match(line):
            in_binary_section = False
            i += 1
            continue
        if git_diff_line.match(line):
            if i + 2 < len(lines) and binary_files_differ.match(lines[i + 2]):
                i += 3
                continue
        if not in_binary_section:
            filtered_lines.append(line)
        i += 1
    filtered_patch = ''.join(filtered_lines)
    return filtered_patch


# Create a sample patch string with binary and text content
sample_patch = """diff --git a/file.txt b/file.txt
index abc1234..def5678 100644
--- a/file.txt
+++ b/file.txt
@@ -1,3 +1,3 @@
 line 1
-line 2
+line 2 modified
 line 3
diff --git a/image.png b/image.png
index 1234567..abcdefg 100644
Binary files a/image.png and b/image.png differ
diff --git a/data.bin b/data.bin
index xyz9999..abc1111 100644
GIT binary patch
literal 42
zcmeAS@N?(olHy`uVBq!ia0vp0!0b-#eH|GXF003eK003eK003eK003eK003eK003eK
diff --git a/normal.py b/normal.py
index 0000000..1111111 100644
--- a/normal.py
+++ b/normal.py
@@ -1 +1,2 @@
 def hello():
+    pass
"""

ARGS = {
    'patch': sample_patch
}
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
format_patch Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/git/utils.py
code served (permissive licence) · get_code("e16f1ea45d6f67d6")
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 logging
import re
from typing import List, Literal, Optional, Union, Tuple, Callable

# Create a simple patch string with various components
patch_content = """diff --git a/file1.txt b/file1.txt
index 1234567..abcdefg 100644
--- a/file1.txt
+++ b/file1.txt
@@ -1,3 +1,3 @@
 line 1
-line 2
+line 2 modified
 line 3
diff --git a/image.bin b/image.bin
index 2345678..bcdefgh 100644
Binary files a/image.bin and b/image.bin differ
diff --git a/file2.txt b/file2.txt
index 3456789..cdefghi 100644
--- a/file2.txt
+++ b/file2.txt
@@ -1,2 +1,2 @@
 content
-old
+new
GIT binary patch
literal 42
zcmeAS@N?(olHy`uVBq!ia0vp0j3CU&3?xW&E$W@V#0NS_^R}7E$M}P9c8l}z3f2~P
z#v?(}l5w(7O%e4E!0sz@E@c^Ekg0$!e&lV=M;Z&XVlCaKVlCaKVlCaKVlCaKVlCaK
VlCaKVlCaKVlCaKVlCaKVlCaKVlCaKVlCaKVlCaKVlCaKVlCaK00000
"""

ARGS = {
    "patch": patch_content
}
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
insert_secret Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/testbeds/utils.py
code served (permissive licence) · get_code("0751670779a93959")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

# Define a simple test script with sys.exit(0)
test_script = """def main():
    print("Running tests...")
    sys.exit(0)
"""

secret = "SECRET_FLAG_12345"

ARGS = {
    "test_script": test_script,
    "secret": secret,
}
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
log_run_subprocess Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/utils.py
code served (permissive licence) · get_code("468a2041e486fed6")
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 coroutine. 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 asyncio
import logging
import sys

# Configure logging to see debug output
logging.basicConfig(level=logging.DEBUG, stream=sys.stdout)

# Define the arguments for log_run_subprocess
ARGS = {
    "command": ["echo", "Hello, World!"],
    "timeout": 5,
    "run_kwargs": {}
}
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 coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine 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
log_run_subprocess_sync Ran
ran; our contract draft was wrong, not the code
qurrent-ai/res-q/resq/utils.py
code served (permissive licence) · get_code("1fda66523279af5a")
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 tuple, 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 logging
import subprocess
import time
from typing import Dict, List, Optional, Tuple, Union, AsyncGenerator, Any

logging.basicConfig(level=logging.DEBUG)

ARGS = {
    "command": ["echo", "hello", "world"],
    "timeout": 10,
}
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 tuple, ndim 1, dtype family float — a tuple 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

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 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 6 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

The instruction-following ability of Large Language Models (LLMs) has cultivated a class of LLM-based systems capable of approaching complex tasks such as making edits to large code repositories. Due to the high sensitivity and unpredictability of LLM behavior in response to changes in prompting, robust evaluation tools are needed to drive future iteration of these systems. We propose RES-Q, a natural language instruction-based benchmark for evaluating $\textbf{R}$epository $\textbf{E}$diting $\textbf{S}$ystems, which consists of 100 handcrafted repository editing tasks derived from real GitHub commits. Given an edit instruction and a code repository, RES-Q evaluates an LLM system's ability to interpret the instruction, navigate the repository to gather relevant information, and construct an appropriate edit that satisfies the specified criteria. We argue that evaluating LLMs in this way addresses issues with traditional benchmarks and provides a more holistic assessment of a model's abilities. We evaluate various state-of-the-art LLMs as language agents in a repository-editing system built on Qurrent OS, our language agent development software. Despite their 1% pass@1 performance difference on HumanEval, we find Claude Sonnet 3.5 outperforms GPT-4o by 12% pass@1 on RES-Q, indicating RES-Q's capacity to differentiate model capability as traditional benchmarks approach saturation. We further investigate token efficiency, performance relationships with existing benchmarks, and interesting disparities between closed and open-source LLMs. Code and dataset are available at https://github.com/Qurrent-AI/RES-Q.

For agents

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

get_harvested_code_for_paper("2406.16801")
get_code_for_paper("2406.16801")
have("2406.16801")

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

curl https://syntology.ai/api/ran/2406.16801.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.16801.svg)](https://syntology.ai/paper/2406.16801)

Connect an agent — have() is free.