SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2306.01102 · 2023

LLMatic: Neural Architecture Search via Large Language Models and Quality Diversity Optimization

arXiv · PDF · Open in the Atlas

Code that ran

11 of 13 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 11 have no contract verdict; 0 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/2306.01102.json · badge: /api/ran/2306.01102.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
umair-nasir14/llmatic canonical (classifier label)
classifier basis: census
11 ran · 2 no recorded run
FunctionStatusWhere it lives
bandit Ran
ran on a synthesized input
umair-nasir14/llmatic/map_elites/multitask.py
code served (permissive licence) · get_code("78bc6168812a4dea")
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 math
import numpy as np
import random

random.seed(42)
np.random.seed(42)

# The function expects:
# - successes: a dict where keys are integers and values are lists of tuples/lists with at least one element
# - n_niches: an integer

# Build successes dict with some sample data
successes = {
    1: [[0.8], [0.7], [0.9]],
    10: [[0.75], [0.85], [0.95], [0.65]],
    50: [[0.88], [0.92], [0.78]],
    100: [[0.91], [0.87], [0.93], [0.89]],
    500: [[0.94], [0.96], [0.85], [0.90], [0.88]]
}

n_niches = 5

ARGS = {
    "successes": successes,
    "n_niches": n_niches
}
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 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
count_parameters Ran
ran on a synthesized input
umair-nasir14/llmatic/utils.py
code served (permissive licence) · get_code("5f4c5fe8e317d932")
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 torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

ARGS = {
    "model": model
}
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 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
get_max_curiosity Ran
ran on a synthesized input
umair-nasir14/llmatic/utils.py
code served (permissive licence) · get_code("3ebacc35fe59b6df")
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

class Species:
    def __init__(self, name, curiosity):
        self.name = name
        self.curiosity = curiosity

species_list = [
    Species("Lion", 7.5),
    Species("Elephant", 8.2),
    Species("Monkey", 9.1),
    Species("Sloth", 3.5),
    Species("Dolphin", 9.8),
]

n = 3

ARGS = {
    "species_list": species_list,
    "n": n,
}
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 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
get_network_width_depth_ratio Ran
ran on a synthesized input
umair-nasir14/llmatic/utils.py
code served (permissive licence) · get_code("d467ccf630194685")
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

torch.manual_seed(0)

# Create a simple neural network for testing
class SimpleNet(nn.Module):
    def __init__(self):
        super(SimpleNet, self).__init__()
        self.conv1 = nn.Conv2d(3, 16, kernel_size=3)
        self.bn1 = nn.BatchNorm2d(16)
        self.relu = nn.ReLU()
        self.fc1 = nn.Linear(16, 32)
        self.fc2 = nn.Linear(32, 10)
    
    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        return x

net = SimpleNet()

ARGS = {
    'net': net
}
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
is_trainable Ran
ran on a synthesized input
umair-nasir14/llmatic/evaluations.py
code served (permissive licence) · get_code("529a95ce1c0c9d06")
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 torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

class Net(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 1, 1, bias=False)
        self.conv2 = nn.Conv2d(3, 1, 1, bias=False)
        self.fc1 = nn.Linear(1024, 10)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        x = F.relu(self.conv2(x))
        x = torch.flatten(x, 1)
        x = F.relu(self.fc1(x))
        return x

net = Net()

ARGS = {
    "net": net
}
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
iso_dd Ran
ran on a synthesized input
umair-nasir14/llmatic/map_elites/common.py
code served (permissive licence) · get_code("3a3a71c514ef66e8")
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 (4,) 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

np.random.seed(0)

# Create small test vectors
x = np.array([1.0, 2.0, 3.0, 4.0])
y = np.array([1.5, 2.5, 3.5, 4.5])

# Create params dict with required keys
params = {
    'max': [10.0, 10.0, 10.0, 10.0],
    'min': [0.0, 0.0, 0.0, 0.0],
    'iso_sigma': 0.1,
    'line_sigma': 0.05
}

ARGS = {
    'x': x,
    'y': y,
    'params': params
}
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 (4,) — shape (4,) 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_centroids Ran
ran on a synthesized input
umair-nasir14/llmatic/plot/plot_2d_map.py
code served (permissive licence) · get_code("f001aeb47ed1c4c4")
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 2, dtype family float, shape (3, 4) 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
import tempfile
import os

# Create a temporary file with centroid data
temp_fd, temp_path = tempfile.mkstemp(suffix='.txt')
try:
    # Write sample centroid data to the temporary file
    centroid_data = np.array([
        [1.0, 2.0, 3.0, 4.0],
        [5.0, 6.0, 7.0, 8.0],
        [9.0, 10.0, 11.0, 12.0]
    ])
    np.savetxt(temp_path, centroid_data)
    
    ARGS = {
        'filename': temp_path
    }
finally:
    os.close(temp_fd)
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 2, dtype family float, shape (3, 4) — shape (3, 4) 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_data Ran
ran on a synthesized input
umair-nasir14/llmatic/plot/plot_2d_map.py
code served (permissive licence) · get_code("93c341f57f1b10fd")
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. 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 tempfile
import os

# Create a temporary file with test data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, 'test_data.txt')

# Create sample data: fit (1 col) + desc (dim cols) + x (dim_x cols)
# For dim=4, dim_x=3: total columns = 1 + 4 + 3 = 8
dim = 4
dim_x = 3
num_samples = 2

# Create data array with shape (num_samples, 1 + dim + dim_x)
data = np.random.randn(num_samples, 1 + dim + dim_x)
np.savetxt(temp_file, data)

ARGS = {
    'filename': temp_file,
    'dim': dim,
    'dim_x': dim_x
}
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 tuple — a tuple output has no array shape of its own; none is 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
polynomial_mutation Ran
ran on a synthesized input
umair-nasir14/llmatic/map_elites/common.py
code served (permissive licence) · get_code("73bf5b8fb69d9f1a")
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 (5,) 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 math
import numpy as np

# Set seed for reproducibility
np.random.seed(42)

# Create a small numpy array as input
x = np.array([0.1, 0.2, 0.3, 0.4, 0.5])

# Define ARGS dict mapping every parameter to a value
ARGS = {
    "x": x
}
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 (5,) — shape (5,) 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
replace_word_mutation Ran
ran on a synthesized input
umair-nasir14/llmatic/mutation_models.py
code served (permissive licence) · get_code("65d84b51afb19593")
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

def replace_word_mutation(sentence):
    if 'Add' in sentence:
        return sentence.replace('Add', 'Delete')
    elif 'Delete' in sentence:
        return sentence.replace('Delete', 'Add')
    else:
        return sentence

ARGS = {
    'sentence': 'Add new feature to the system'
}
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
select_niche Ran
ran on a synthesized input
umair-nasir14/llmatic/map_elites/multitask.py
code served (permissive licence) · get_code("e111d04d93bb13e1")
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. 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
from scipy.spatial import distance

class DummyX:
    def __init__(self):
        self.centroid = np.array([0.1, 0.2, 0.3, 0.4])

np.random.seed(0)

x = DummyX()

z = np.array([0.5, 0.6, 0.7, 0.8])

f = 1.23

centroids = np.array([
    [0.1, 0.2, 0.3, 0.4],
    [0.9, 0.8, 0.7, 0.6],
    [0.2, 0.3, 0.4, 0.5],
    [0.4, 0.5, 0.6, 0.7]
])

tasks = [
    {"name": "task_0", "value": 10},
    {"name": "task_1", "value": 20},
    {"name": "task_2", "value": 30},
    {"name": "task_3", "value": 40}
]

t_size = 2

params = {"lr": 0.01, "epochs": 100, "batch_size": 32}

use_distance = False

ARGS = {
    "x": x,
    "z": z,
    "f": f,
    "centroids": centroids,
    "tasks": tasks,
    "t_size": t_size,
    "params": params,
    "use_distance": use_distance
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
sbx Not yet run
unverified (harvested, no recorded run)
umair-nasir14/llmatic/map_elites/common.py
code served (permissive licence) · get_code("3c6472fd840d3195")
How it ran

no invocation recorded.

voronoi_finite_polygons_2d Not yet run
unverified (harvested, no recorded run)
umair-nasir14/llmatic/plot/plot_2d_map.py
code served (permissive licence) · get_code("230de062c09a7ba5")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 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 13 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 (LLMs) have emerged as powerful tools capable of accomplishing a broad spectrum of tasks. Their abilities span numerous areas, and one area where they have made a significant impact is in the domain of code generation. Here, we propose using the coding abilities of LLMs to introduce meaningful variations to code defining neural networks. Meanwhile, Quality-Diversity (QD) algorithms are known to discover diverse and robust solutions. By merging the code-generating abilities of LLMs with the diversity and robustness of QD solutions, we introduce \texttt{LLMatic}, a Neural Architecture Search (NAS) algorithm. While LLMs struggle to conduct NAS directly through prompts, \texttt{LLMatic} uses a procedural approach, leveraging QD for prompts and network architecture to create diverse and high-performing networks. We test \texttt{LLMatic} on the CIFAR-10 and NAS-bench-201 benchmarks, demonstrating that it can produce competitive networks while evaluating just $2,000$ candidates, even without prior knowledge of the benchmark domain or exposure to any previous top-performing models for the benchmark. The open-sourced code is available in \url{https://github.com/umair-nasir14/LLMatic}.

For agents

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

get_harvested_code_for_paper("2306.01102")
get_code_for_paper("2306.01102")
have("2306.01102")

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

curl https://syntology.ai/api/ran/2306.01102.json

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

[![Syntology run record](https://syntology.ai/api/ran/2306.01102.svg)](https://syntology.ai/paper/2306.01102)

Connect an agent — have() is free.