SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.01098 · ICLR · 2022

Evaluation Metrics for Graph Generative Models: Problems, Pitfalls, and Practical Solutions

Bastian Rieck, Karsten Borgwardt, Leslie O'bray, Max Horn

arXiv · PDF · Open in the Atlas

Code that ran

12 of 15 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 12 have no contract verdict; 0 failures of our instrument; 3 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/2106.01098.json · badge: /api/ran/2106.01098.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
borgwardtlab/ggme canonical (classifier label)
classifier basis: census
12 ran · 3 no recorded run
FunctionStatusWhere it lives
adj_to_networkx Ran
ran on a synthesized input
borgwardtlab/ggme/src/utils.py
code served (permissive licence) · get_code("6c1cf18a77ad5c33")
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 networkx as nx
import numpy as np

# Create small adjacency matrices as numpy arrays
# The function expects an iterable of adjacency matrices
adj_matrix_1 = np.array([
    [0, 1, 0, 1],
    [1, 0, 1, 0],
    [0, 1, 0, 1],
    [1, 0, 1, 0]
], dtype=float)

adj_matrix_2 = np.array([
    [0, 1, 1],
    [1, 0, 1],
    [1, 1, 0]
], dtype=float)

graphs = [adj_matrix_1, adj_matrix_2]

ARGS = {
    "graphs": graphs
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, numpy — 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
clustering_coefficient Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/descriptor_functions.py
code served (permissive licence) · get_code("e69b3f5ed74f8a34")
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 int, shape (200,) 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 networkx as nx

# Create a simple graph for testing
G = nx.Graph()
G.add_edges_from([(0, 1), (1, 2), (2, 0), (2, 3), (3, 4), (4, 5)])

# Define arguments dict
ARGS = {
    'G': G,
    'n_bins': 200,
    'density': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, 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 int, shape (200,) — shape (200,) 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
compute_correlation Ran
ran on a synthesized input
borgwardtlab/ggme/src/correlation.py
code served (permissive licence) · get_code("09d321100886c8b5")
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 float64, ndim 0, 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
import scipy as sp

torch.manual_seed(0)

# Create perturbation_values: a list of perturbation levels
perturbation_values = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5]

# Create list_of_mmd: a list of MMD (Maximum Mean Discrepancy) values
# These should correlate with perturbation_values for a meaningful test
list_of_mmd = [1.0, 1.05, 1.15, 1.25, 1.35, 1.5]

# correlation_type: specify the type of correlation to compute
correlation_type = 'pearson'

ARGS = {
    'perturbation_values': perturbation_values,
    'list_of_mmd': list_of_mmd,
    'correlation_type': correlation_type
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
degree_distribution Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/descriptor_functions.py
code served (permissive licence) · get_code("5a0fbb727abb2bb5")
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 int, 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
import networkx as nx

# Create a simple graph for testing
G = nx.Graph()
G.add_edges_from([(0, 1), (1, 2), (2, 3), (3, 0), (1, 3)])

# Define arguments dict
ARGS = {
    "G": G,
    "density": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, 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 int, 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
ensure_padded Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/utils.py
code served (permissive licence) · get_code("923f7e15bd130842")
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 3, 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 numpy as np

# Create test data for ensure_padded function
# X is a list of arrays with different lengths
X = [
    np.array([1.0, 2.0, 3.0]),
    np.array([4.0, 5.0]),
    np.array([6.0, 7.0, 8.0, 9.0])
]

# Y is a list of arrays with different lengths
Y = [
    np.array([10.0, 11.0]),
    np.array([12.0, 13.0, 14.0]),
    np.array([15.0])
]

# Test case 1: Both X and Y provided
ARGS = {
    'X': X,
    'Y': Y
}
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, ndim 3, 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
gaussian_kernel Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/kernels.py
code served (permissive licence) · get_code("31a99a406af1fcf6")
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 float64, ndim 0, 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 numpy as np

# The function takes x, y as numpy arrays and computes a Gaussian kernel
# It performs: np.exp(-sigma * np.dot(x - y, x - y))
# x and y should be 1D arrays of the same shape

x = np.array([1.0, 2.0, 3.0, 4.0])
y = np.array([1.5, 2.5, 3.5, 4.5])
sigma = 1.0

ARGS = {
    "x": x,
    "y": y,
    "sigma": sigma,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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_open_fn Ran
ran on a synthesized input
borgwardtlab/ggme/src/utils.py
code served (permissive licence) · get_code("72eed74671718b01")
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 builtin_function_or_method. 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 gzip
import bz2

# The function get_open_fn takes a filepath string and returns the appropriate
# open function based on the file extension. We need to test it with different
# file paths.

ARGS = {
    'filepath': 'example.txt'
}
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 builtin_function_or_method — a builtin_function_or_method output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a builtin_function_or_method 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
laplacian_total_variation_kernel Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/kernels.py
code served (permissive licence) · get_code("637718871ee800c0")
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 float64, ndim 0, 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 numpy as np

# Set up arguments for laplacian_total_variation_kernel
x = np.array([1.0, 2.0, 3.0, 4.0])
y = np.array([1.5, 2.5, 3.5, 4.5])
sigma = 1.0

ARGS = {
    "x": x,
    "y": y,
    "sigma": sigma
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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_graphs Ran
ran on a synthesized input
borgwardtlab/ggme/src/utils.py
code served (permissive licence) · get_code("7253629098c52c80")
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 2, 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 gzip
import bz2
import pickle
import tempfile
import os
from scipy.sparse import csr_matrix
import numpy as np
import networkx as nx

# Create a temporary directory for test files
temp_dir = tempfile.mkdtemp()

# Test case 1: Create graphs as numpy arrays and save to pickle file
np.random.seed(0)
graphs_np = [np.array([[0, 1, 0], [1, 0, 1], [0, 1, 0]]), 
             np.array([[0, 1, 1], [1, 0, 0], [1, 0, 0]])]
filepath_np = os.path.join(temp_dir, "graphs_np.pkl")
with open(filepath_np, 'wb') as f:
    pickle.dump(graphs_np, f)

# Test case 2: Create graphs as sparse matrices and save to compressed pickle file
graphs_sparse = [csr_matrix(np.array([[0, 1, 0], [1, 0, 1], [0, 1, 0]])),
                 csr_matrix(np.array([[0, 1, 1], [1, 0, 0], [1, 0, 0]]))]
filepath_sparse_gz = os.path.join(temp_dir, "graphs_sparse.pkl.gz")
with gzip.open(filepath_sparse_gz, 'wb') as f:
    pickle.dump(graphs_sparse, f)

# Test case 3: Create networkx graphs and save to bz2 compressed pickle file
graphs_nx = [nx.Graph([(0, 1), (1, 2)]), 
             nx.Graph([(0, 1), (0, 2)])]
filepath_nx_bz2 = os.path.join(temp_dir, "graphs_nx.pkl.bz2")
with bz2.open(filepath_nx_bz2, 'wb') as f:
    pickle.dump(graphs_nx, f)

# Define ARGS with the filepath as the parameter
ARGS = {
    'filepath': filepath_np
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, 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
mmd_variance_estimate Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/mmd.py
code served (permissive licence) · get_code("17653f52e7cb710a")
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 float64, ndim 0, 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 numpy as np

# Create small kernel matrices
# K_XX: (m, m) - kernel between samples from first distribution
# K_YY: (n, n) - kernel between samples from second distribution
# K_XY: (m, n) - kernel between samples from both distributions
# Requirement: m == n

m = 3
n = 3

# Create symmetric positive semi-definite kernel matrices
np.random.seed(0)

# K_XX: (3, 3) kernel matrix
K_XX_temp = np.random.randn(m, m)
K_XX = (K_XX_temp + K_XX_temp.T) / 2 + np.eye(m)

# K_YY: (3, 3) kernel matrix
K_YY_temp = np.random.randn(n, n)
K_YY = (K_YY_temp + K_YY_temp.T) / 2 + np.eye(n)

# K_XY: (3, 3) kernel matrix between the two distributions
K_XY = np.random.randn(m, n)

# Default lambda parameter
lam = 1e-08

ARGS = {
    "K_XX": K_XX,
    "K_YY": K_YY,
    "K_XY": K_XY,
    "lam": lam
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
normalised_laplacian_spectrum Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/descriptor_functions.py
code served (permissive licence) · get_code("a401d35796446a06")
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 int, shape (200,) 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 networkx as nx

# Create a small graph for testing
G = nx.karate_club_graph()

# Define ARGS dict mapping all parameters
ARGS = {
    "G": G,
    "n_bins": 200,
    "bin_range": (0, 2),
    "density": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, 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 int, shape (200,) — shape (200,) 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
pad_to_length Ran
ran on a synthesized input
borgwardtlab/ggme/src/metrics/utils.py
code served (permissive licence) · get_code("323380a27091cb1b")
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 int, shape (7,) 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 small numpy array to pad
element = np.array([1, 2, 3])

# Define the target length for padding
length = 7

# Construct the arguments dictionary
ARGS = {
    "element": element,
    "length": length
}
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 int, shape (7,) — shape (7,) 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
gaussian_tv Not yet run
unverified (harvested, no recorded run)
borgwardtlab/ggme/src/metrics/kernels.py
code served (permissive licence) · get_code("3996e1aecc59e905")
How it ran

no invocation recorded.

mmd Not yet run
unverified (harvested, no recorded run)
borgwardtlab/ggme/src/metrics/mmd.py
code served (permissive licence) · get_code("b8ddf9e892aa4b13")
How it ran

no invocation recorded.

mmd_linear_approximation Not yet run
unverified (harvested, no recorded run)
borgwardtlab/ggme/src/metrics/mmd.py
code served (permissive licence) · get_code("fcd59998ef67f1a7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 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 15 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

Graph generative models are a highly active branch of machine learning. Given the steady development of new models of ever-increasing complexity, it is necessary to provide a principled way to evaluate and compare them. In this paper, we enumerate the desirable criteria for such a comparison metric and provide an overview of the status quo of graph generative model comparison in use today, which predominantly relies on the maximum mean discrepancy (MMD). We perform a systematic evaluation of MMD in the context of graph generative model comparison, highlighting some of the challenges and pitfalls researchers inadvertently may encounter. After conducting a thorough analysis of the behaviour of MMD on synthetically-generated perturbed graphs as well as on recently-proposed graph generative models, we are able to provide a suitable procedure to mitigate these challenges and pitfalls. We aggregate our findings into a list of practical recommendations for researchers to use when evaluating graph generative models.

For agents

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

get_harvested_code_for_paper("2106.01098")
get_code_for_paper("2106.01098")
have("2106.01098")

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

curl https://syntology.ai/api/ran/2106.01098.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.01098.svg)](https://syntology.ai/paper/2106.01098)

Connect an agent — have() is free.