SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.17396 · NeurIPS · 2024

Fairness-Aware Estimation of Graphical Models

Li Shen, Qi Long, Zhuoping Zhou, Davoud Tarzanagh, Bojian Hou

arXiv · PDF · Open in the Atlas

Code that ran

14 of 14 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; 4 honoured the contract we drafted, 3 violated it and 0 have no contract verdict; 7 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/2408.17396.json · badge: /api/ran/2408.17396.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
pennshenlab/fair_gms canonical (classifier label)
classifier basis: census
14 ran · 0 no recorded run
FunctionStatusWhere it lives
HubNetwork Ran
ran; our contract draft was wrong, not the code
pennshenlab/fair_gms/GRAPH_Framework-main/utils/common.py
code served (permissive licence) · get_code("b96e101dc2501196")
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 dict. 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

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

ARGS = {
    'p': 5,
    'sparsity': 0.3,
    'hubnumber': 2,
    'hubsparsity': 0.4,
    'type': 'Gaussian',
    'hubcol': None
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
binary_mcmc Ran
ran; our contract draft was wrong, not the code
pennshenlab/fair_gms/GRAPH_Framework-main/utils/common.py
code served (permissive licence) · get_code("f3a12f6c8cc6a022")
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 (10, 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)

# Construct arguments for binary_mcmc
n = 10  # number of samples to draw
p = 4   # number of dimensions
Theta = np.random.randn(p, p)  # p x p parameter matrix
burnin = 100  # burnin iterations
skip = 5  # skip rate for sampling
trace = False  # whether to print trace

ARGS = {
    "n": n,
    "Theta": Theta,
    "burnin": burnin,
    "skip": skip,
    "trace": trace,
}
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 (10, 4) — shape (10, 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
loss Ran
ran, violated the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/covariance.py
code served (permissive licence) · get_code("d4ca96158400da5d")
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 square matrices for the loss function
# A and S should be square matrices (required for determinant and Frobenius norm)
np.random.seed(0)

# Create a 4x4 matrix A
A = np.random.randn(4, 4)

# Create a 4x4 matrix S with similar shape
S = np.random.randn(4, 4)

# tau should be a positive scalar (used as a coefficient in the log term)
tau = 0.1

ARGS = {
    "A": A,
    "S": S,
    "tau": tau
}
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
loss Ran
ran, honoured the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/ising.py
code served (permissive licence) · get_code("ac2bd3f200c6567a")
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

# Construct valid arguments for the loss function
# The function expects:
# - Theta: a p x p matrix (parameter matrix)
# - X: an n x p matrix (data matrix)

# Set dimensions: n=2 (samples), p=4 (features)
n = 2
p = 4

# Create X as an n x p matrix
X = np.random.randn(n, p)

# Create Theta as a p x p matrix
Theta = np.random.randn(p, p)

ARGS = {
    "Theta": Theta,
    "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 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
loss Ran
ran, honoured the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/precision.py
code served (permissive licence) · get_code("afb898f1b9f28dda")
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

# Construct valid arguments for the loss function
# The function takes two matrices A and S
# It computes: -log(det(A)) + trace(S @ A)
# Both should be square matrices for det and trace to work

np.random.seed(0)

# Create a small 4x4 positive definite matrix A
# (positive definite ensures det(A) > 0 for valid log)
A_base = np.random.randn(4, 4)
A = A_base @ A_base.T + np.eye(4)  # Make it positive definite

# Create a small 4x4 matrix S
S = np.random.randn(4, 4)

ARGS = {
    "A": A,
    "S": S,
}
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
loss_l1 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/covariance.py
code served (permissive licence) · get_code("69334a9edd068329")
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 random seed for reproducibility
np.random.seed(0)

# Create small matrices for testing
# A is a square matrix (the main variable)
# S is a sample covariance matrix (same shape as A)
# tau is a scalar parameter
# lam is a scalar regularization parameter
# off_diag is a boolean flag

# Create a 4x4 positive definite matrix for A (symmetric, required for det)
A = np.array([
    [2.0, 0.5, 0.1, 0.2],
    [0.5, 1.8, 0.3, 0.1],
    [0.1, 0.3, 2.2, 0.4],
    [0.2, 0.1, 0.4, 1.9]
])

# Create a 4x4 positive definite matrix for S
S = np.array([
    [1.9, 0.4, 0.05, 0.15],
    [0.4, 1.7, 0.2, 0.05],
    [0.05, 0.2, 2.1, 0.3],
    [0.15, 0.05, 0.3, 1.8]
])

# Scalar parameter tau
tau = 0.5

# Scalar regularization parameter lam
lam = 0.1

# Boolean flag for off-diagonal penalty
off_diag = False

ARGS = {
    'A': A,
    'S': S,
    'tau': tau,
    'lam': lam,
    'off_diag': off_diag
}
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
loss_l1 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/ising.py
code served (permissive licence) · get_code("199618a5b413eb92")
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

# Construct small test data
np.random.seed(0)

# X should be n x p matrix
# The function does: X.T @ X, X.T, etc.
# Let's use n=2 (samples), p=4 (features)
n = 2
p = 4
X = np.random.randn(n, p)

# Theta should be p x p matrix (precision matrix)
# The function does: Theta @ X.T, diag operations on Theta, etc.
Theta = np.random.randn(p, p)

# Lambda is a scalar regularization parameter
lam = 0.1

# off_diag is a boolean flag
off_diag = False

ARGS = {
    "Theta": Theta,
    "X": X,
    "lam": lam,
    "off_diag": off_diag
}
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
loss_l1 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/precision.py
code served (permissive licence) · get_code("6c8f49e52fde5cfd")
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 seed for reproducibility
np.random.seed(0)

# Create small square matrices for A and S
# A should be invertible (positive definite) for det(A) to be positive
# S is used in trace(S @ A), so both should be square matrices
dim = 4
A = np.random.randn(dim, dim)
# Make A positive definite by A = A^T @ A
A = A.T @ A

S = np.random.randn(dim, dim)

# Lambda parameter for L1 regularization
lam = 0.1

# off_diag parameter to control regularization type
off_diag = False

# Construct ARGS dict with all parameters
ARGS = {
    "A": A,
    "S": S,
    "lam": lam,
    "off_diag": off_diag,
}
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
minimize_proximal_gradient Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pennshenlab/fair_gms/GRAPH_Framework-main/utils/proximal_gradient.py
code served (permissive licence) · get_code("44d40136fd1be1b1")
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 OptimizeResult. 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.optimize import OptimizeResult

# Set random seed for reproducibility
np.random.seed(0)

# Define objective function f (multi-objective, returns vector)
def f(x):
    # Returns 2 objectives
    return np.array([
        np.sum(x**2),
        np.sum((x - 1)**2)
    ])

# Define regularization function g
def g(x):
    # L1 norm regularization
    return np.sum(np.abs(x))

# Define Jacobian of f
def jac_f(x):
    # Returns shape (n_objectives, *x.shape)
    n_objectives = 2
    jac = np.zeros((n_objectives, *x.shape))
    jac[0] = 2 * x  # gradient of sum(x^2)
    jac[1] = 2 * (x - 1)  # gradient of sum((x-1)^2)
    return jac

# Define proximal operator of weighted sum of g
def prox_wsum_g(x, lam):
    # Proximal operator of L1 norm: soft thresholding
    # prox_{lambda*||.||_1}(x) = sign(x) * max(|x| - lambda, 0)
    return np.sign(x) * np.maximum(np.abs(x) - lam, 0)

# Initial point
x0 = np.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    'f': f,
    'g': g,
    'jac_f': jac_f,
    'prox_wsum_g': prox_wsum_g,
    'x0': x0,
    'lr': 0.01,
    'lam': 0.01,
    'tol': 1e-5,
    'tol_internal': 1e-12,
    'max_iter': 100,
    'max_iter_internal': 1000,
    'max_backtrack_iter': 100,
    'warm_start': False,
    'decay_rate': 0.1,
    'nesterov': False,
    'nesterov_ratio': (0, 0.25),
    'return_all': False,
    'verbose': False,
    'deprecated': False
}
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 OptimizeResult — a OptimizeResult output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a OptimizeResult 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
objective_Q Ran
ran, violated the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/algos/GRAPH/Covariance.py
code served (permissive licence) · get_code("8ce182b38a53e918")
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

# objective_Q expects:
# - objective_f_value: a scalar (float or numpy scalar)
# - A: a 2D array for matrix operations (@ and -)
# - D: a 2D array for matrix operations (@ and -)
# - A_next: a 2D array for matrix operations (@ and -)
# - step: a scalar (int or float) for division

# The function computes:
# A_next_A = A_next - A  (matrix subtraction)
# np.trace(A_next_A @ D)  (matrix multiplication, then trace)
# np.sum(np.square(A_next_A))  (element-wise square, then sum)

# For valid matrix operations, dimensions must be compatible:
# A_next - A requires same shape
# A_next_A @ D requires (m, n) @ (n, p) -> (m, p), but trace only works on square
# So A_next_A must be square, and D must be square with compatible dimensions

np.random.seed(0)

# Create square matrices of size 4x4
A = np.random.randn(4, 4)
A_next = np.random.randn(4, 4)
D = np.random.randn(4, 4)

# Scalar value for objective function
objective_f_value = 1.5

# Step size (positive scalar)
step = 0.01

ARGS = {
    "objective_f_value": objective_f_value,
    "A": A,
    "D": D,
    "A_next": A_next,
    "step": step,
}
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
objective_f_grad Ran
ran, violated the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/covariance.py
code served (permissive licence) · get_code("ffba8da0183b2eaa")
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 (4, 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

# Construct valid arguments for objective_f_grad
# The function takes three parameters: A, S, tau
# A and S are used in matrix operations (inversion, subtraction)
# tau is a scalar multiplier

np.random.seed(0)

# A: a square invertible matrix (4x4)
A = np.array([[2.0, 0.5, 0.1, 0.2],
              [0.5, 2.5, 0.3, 0.1],
              [0.1, 0.3, 2.2, 0.4],
              [0.2, 0.1, 0.4, 2.1]], dtype=np.float64)

# S: a matrix of same shape as A (4x4)
S = np.array([[1.0, 0.2, 0.1, 0.05],
              [0.2, 1.5, 0.2, 0.1],
              [0.1, 0.2, 1.3, 0.15],
              [0.05, 0.1, 0.15, 1.2]], dtype=np.float64)

# tau: a scalar regularization parameter
tau = 0.5

ARGS = {
    "A": A,
    "S": S,
    "tau": tau
}
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 (4, 4) — shape (4, 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
objective_f_grad Ran
ran, honoured the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/ising.py
code served (permissive licence) · get_code("27681ac3593321ca")
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 (4, 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

# Set random seed for reproducibility
np.random.seed(0)

# Small dimensions: n=2 samples, p=4 parameters
n = 2
p = 4

# Create small test matrices
X = np.random.randn(n, p)
Theta = np.random.randn(p, p)

# Construct the arguments dictionary
ARGS = {
    "Theta": Theta,
    "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 2, dtype family float, shape (4, 4) — shape (4, 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
objective_f_grad Ran
ran, honoured the contract we drafted
pennshenlab/fair_gms/GRAPH_Framework-main/utils/GRAPH/precision.py
code served (permissive licence) · get_code("583941f6d99b7364")
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 Tensor, ndim 2, dtype family float, shape (2, 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 torch
import numpy as np

torch.manual_seed(0)

# The function performs S - A_inv, so both parameters need to be tensors
# that support subtraction and are broadcastable or same shape
A_inv = torch.randn(2, 4)
S = torch.randn(2, 4)

ARGS = {
    "A_inv": A_inv,
    "S": S,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
soft_threshold Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pennshenlab/fair_gms/GRAPH_Framework-main/utils/common.py
code served (permissive licence) · get_code("0e13dc21d49f0caf")
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 (4, 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

# Create a small matrix to test soft_threshold
X = np.array([
    [0.5, -0.3, 0.8, -0.1],
    [-0.2, 0.9, -0.4, 0.6],
    [0.7, -0.5, 0.2, -0.3],
    [-0.6, 0.4, -0.1, 0.8]
], dtype='float32')

threshold = 0.3

modify_diag = False

ARGS = {
    'X': X,
    'threshold': threshold,
    'modify_diag': modify_diag
}
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 (4, 4) — shape (4, 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

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 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 14 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

This paper examines the issue of fairness in the estimation of graphical models (GMs), particularly Gaussian, Covariance, and Ising models. These models play a vital role in understanding complex relationships in high-dimensional data. However, standard GMs can result in biased outcomes, especially when the underlying data involves sensitive characteristics or protected groups. To address this, we introduce a comprehensive framework designed to reduce bias in the estimation of GMs related to protected attributes. Our approach involves the integration of the pairwise graph disparity error and a tailored loss function into a nonsmooth multi-objective optimization problem, striving to achieve fairness across different sensitive groups while maintaining the effectiveness of the GMs. Experimental evaluations on synthetic and real-world datasets demonstrate that our framework effectively mitigates bias without undermining GMs' performance.

For agents

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

get_harvested_code_for_paper("2408.17396")
get_code_for_paper("2408.17396")
have("2408.17396")

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

curl https://syntology.ai/api/ran/2408.17396.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.17396.svg)](https://syntology.ai/paper/2408.17396)

Connect an agent — have() is free.