SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1902.09635 · 2019

NAS-Bench-101: Towards Reproducible Neural Architecture Search

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 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 9 have no contract verdict; 0 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/1902.09635.json · badge: /api/ran/1902.09635.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
DaveKim3872/nasbench-hpc community (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
FunctionStatusWhere it lives
gen_is_edge_fn Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/nasbench/lib/graph_util.py
code served (permissive licence) · get_code("b558868dfd1955d6")
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 vectorize. 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

# The function gen_is_edge_fn takes a single parameter: bits
# bits should be an integer that will be interpreted as a bit mask
# Let's use a small integer that will create some edges in the adjacency pattern

ARGS = {
    "bits": 21  # Binary: 010101, which sets bits at positions 0, 2, 4
}
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 vectorize — a vectorize output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a vectorize 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
is_full_dag Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/nasbench/lib/graph_util.py
code served (permissive licence) · get_code("283b1ed886dfa4c1")
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

# Create a valid full DAG matrix
# A full DAG means:
# 1) All rows except the last have at least one 1 (all vertices have outgoing edges)
# 2) All columns except the first have at least one 1 (all vertices have incoming edges)
# 3) Upper triangular (no cycles)

# 5x5 upper triangular matrix representing a full DAG
matrix = np.array([
    [0, 1, 1, 0, 0],  # vertex 0 has outgoing edges
    [0, 0, 1, 1, 0],  # vertex 1 has outgoing edges
    [0, 0, 0, 1, 1],  # vertex 2 has outgoing edges
    [0, 0, 0, 0, 1],  # vertex 3 has outgoing edges
    [0, 0, 0, 0, 0],  # vertex 4 (output) has no outgoing edges - OK
], dtype=int)

ARGS = {
    "matrix": matrix
}
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
is_included Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/delete_redundancy.py
code served (permissive licence) · get_code("4e196702d29923e4")
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

# Create sample items with module_adjacency and module_operations
item1 = {
    'module_adjacency': [[1, 0, 1], [0, 1, 0], [1, 0, 1]],
    'module_operations': ['conv', 'relu', 'pool']
}

item2 = {
    'module_adjacency': [[1, 1, 0], [1, 0, 1], [0, 1, 1]],
    'module_operations': ['conv', 'batch_norm', 'pool']
}

item3 = {
    'module_adjacency': [[1, 0, 1], [0, 1, 0], [1, 0, 1]],
    'module_operations': ['conv', 'relu', 'pool']
}

item_to_check = {
    'module_adjacency': [[1, 0, 1], [0, 1, 0], [1, 0, 1]],
    'module_operations': ['conv', 'relu', 'pool']
}

item_list = [item1, item2, item3]

ARGS = {
    'item': item_to_check,
    'item_list': item_list
}
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
is_included Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/delete_with_step.py
code served (permissive licence) · get_code("18f8dd429732fc7a")
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 numpy as np

# Create sample item dictionaries with module_adjacency and module_operations
item1 = {
    'module_adjacency': np.array([[0, 1], [1, 0]]),
    'module_operations': np.array([1, 2]),
    'index': 0
}

item2 = {
    'module_adjacency': np.array([[0, 1, 0], [1, 0, 1], [0, 1, 0]]),
    'module_operations': np.array([1, 2, 3]),
    'index': 1
}

item3 = {
    'module_adjacency': np.array([[0, 1], [1, 0]]),
    'module_operations': np.array([1, 2]),
    'index': 2
}

# Create the query item (should match item1 and item3)
query_item = {
    'module_adjacency': np.array([[0, 1], [1, 0]]),
    'module_operations': np.array([1, 2])
}

# Create item_list with multiple items
item_list = [item1, item2, item3]

ARGS = {
    'item': query_item,
    'item_list': item_list
}
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
is_upper_triangular Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/nasbench/lib/model_spec.py
code served (permissive licence) · get_code("de7246ef524e6189")
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

# Create a test matrix that is upper triangular
# (zeros on and below the diagonal)
upper_triangular_matrix = np.array([
    [1, 2, 3, 4],
    [0, 5, 6, 7],
    [0, 0, 8, 9],
    [0, 0, 0, 10]
], dtype=float)

ARGS = {
    "matrix": upper_triangular_matrix
}
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
matrix_eq_comp Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/delete_redundancy.py
code served (permissive licence) · get_code("8912b1df700ddccb")
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 int64, 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 numpy as np

# The function takes two arrays and compares their elements
# It uses np.intersect1d to find common elements
# It uses np.bincount to count occurrences
# So we need 1D integer arrays with some overlapping elements

mat1 = np.array([0, 1, 2, 1, 3, 2, 2], dtype=np.int64)
mat2 = np.array([1, 2, 2, 4, 1, 3, 3], dtype=np.int64)

ARGS = {
    "mat1": mat1,
    "mat2": mat2
}
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 int64, 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
num_edges Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/nasbench/lib/graph_util.py
code served (permissive licence) · get_code("2e655699a68fae2a")
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 int64, 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 numpy as np

# Create a simple adjacency matrix
# The function sums all elements, so we need a numeric matrix
# Using a small 3x3 adjacency matrix with some edges
adjacency_matrix = np.array([
    [0, 1, 1],
    [1, 0, 1],
    [1, 1, 0]
], dtype=np.int32)

ARGS = {
    "matrix": adjacency_matrix
}
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 int64, 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
one_dif Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/delete_redundancy.py
code served (permissive licence) · get_code("8cbe0e9bb117c4c8")
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

# Create two dict items with 'module_adjacency' and 'module_operations' keys
# These dicts represent neural architecture search (NAS) cells

# item0: base configuration
item0 = {
    'module_adjacency': np.array([0, 1, 1, 0, 1, 0]),  # adjacency matrix flattened
    'module_operations': np.array(['conv', 'relu', 'pool'])  # operations
}

# item1: differs by exactly one operation and one adjacency element
item1 = {
    'module_adjacency': np.array([0, 1, 1, 0, 0, 0]),  # differs in one position from item0
    'module_operations': np.array(['conv', 'sigmoid', 'pool'])  # differs in one position from item0
}

ARGS = {
    'item0': item0,
    'item1': item1
}
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
one_dif Ran
ran on a synthesized input
DaveKim3872/nasbench-hpc/delete_with_step.py
code served (permissive licence) · get_code("ef1f44ad9b01eddd")
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

def create_item(num_modules, num_edges, num_operations_different=False):
    """Create a dict item with module_adjacency and module_operations."""
    # Create module_operations: array of operation strings
    operations = np.array(['op_a', 'op_b', 'op_c', 'op_d'][:num_modules], dtype=object)
    
    # Create module_adjacency: flattened adjacency matrix
    adj_size = num_modules * (num_modules - 1) // 2
    adjacency = np.random.randint(0, 2, size=adj_size)
    
    item = {
        'module_adjacency': adjacency,
        'module_operations': operations
    }
    return item

# Create item0: base item
item0 = create_item(num_modules=4, num_edges=6)

# Create item1: differs by exactly 1 operation and 0 adjacencies
# (so count = 1 total, satisfying the function's return True condition)
item1 = create_item(num_modules=4, num_edges=6)
item1['module_operations'] = item0['module_operations'].copy()
item1['module_adjacency'] = item0['module_adjacency'].copy()
# Change exactly one operation
item1['module_operations'][0] = 'op_different'

ARGS = {
    'item0': item0,
    'item1': item1
}
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
conv_bn_relu Not yet run
unverified (harvested, no recorded run)
DaveKim3872/nasbench-hpc/nasbench/lib/base_ops.py
code served (permissive licence) · get_code("1500e49c46a26813")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 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.

FunctionAlso present inEarliest-datedOthers named
conv_bn_relu 1 other paper
1 later
1912.00848 (2019-12)
Neural Predictor for Neural Architecture Search
gen_is_edge_fn 1 other paper
1 later
1912.00848 (2019-12)
Neural Predictor for Neural Architecture Search
is_full_dag 1 other paper
1 later
1912.00848 (2019-12)
Neural Predictor for Neural Architecture Search
num_edges 1 other paper
1 later
1912.00848 (2019-12)
Neural Predictor for Neural Architecture Search

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

Recent advances in neural architecture search (NAS) demand tremendous computational resources, which makes it difficult to reproduce experiments and imposes a barrier-to-entry to researchers without access to large-scale computation. We aim to ameliorate these problems by introducing NAS-Bench-101, the first public architecture dataset for NAS research. To build NAS-Bench-101, we carefully constructed a compact, yet expressive, search space, exploiting graph isomorphisms to identify 423k unique convolutional architectures. We trained and evaluated all of these architectures multiple times on CIFAR-10 and compiled the results into a large dataset of over 5 million trained models. This allows researchers to evaluate the quality of a diverse range of models in milliseconds by querying the pre-computed dataset. We demonstrate its utility by analyzing the dataset as a whole and by benchmarking a range of architecture optimization algorithms.

For agents

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

get_harvested_code_for_paper("1902.09635")
get_code_for_paper("1902.09635")
have("1902.09635")

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

curl https://syntology.ai/api/ran/1902.09635.json

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

[![Syntology run record](https://syntology.ai/api/ran/1902.09635.svg)](https://syntology.ai/paper/1902.09635)

Connect an agent — have() is free.