SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2210.00364 · ICLR · 2023

DCI-ES: An Extended Disentanglement Framework with Connections to Identifiability

Bernhard Schölkopf, Julius Von Kügelgen, Frederik Träuble, Andrea Dittadi, Andrei Nicolicioiu, Cian Eastwood, Armin Kekić

arXiv · PDF · Open in the Atlas

Code that ran

13 of 18 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; 1 honoured the contract we drafted, 4 violated it and 5 have no contract verdict; 3 failures of our instrument; 5 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/2210.00364.json · badge: /api/ran/2210.00364.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
iancovert/sage canonical (classifier label)
classifier basis not recorded on the link
7 ran · 5 no recorded run
andreinicolicioiu/dci-es canonical (classifier label)
classifier basis: classifier_2025
6 ran · 0 no recorded run
FunctionStatusWhere it lives
_compute_dci Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("464d70bba860709c")
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) 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
import scipy
from sklearn import ensemble

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

# Create small synthetic data for the function
# mus_train: shape [num_codes, num_train_samples]
# ys_train: shape [num_factors, num_train_samples]
# mus_test: shape [num_codes, num_test_samples]
# ys_test: shape [num_factors, num_test_samples]

num_codes = 4
num_factors = 3
num_train_samples = 8
num_test_samples = 6

# Generate random continuous codes (latent representations)
mus_train = np.random.randn(num_codes, num_train_samples).astype(np.float64)
mus_test = np.random.randn(num_codes, num_test_samples).astype(np.float64)

# Generate random binary factors (ground truth factors)
ys_train = np.random.randint(0, 2, size=(num_factors, num_train_samples)).astype(np.float64)
ys_test = np.random.randint(0, 2, size=(num_factors, num_test_samples)).astype(np.float64)

ARGS = {
    'mus_train': mus_train,
    'ys_train': ys_train,
    'mus_test': mus_test,
    'ys_test': ys_test,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, scipy, sklearn — 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
calculate_A Ran
ran, honoured the contract we drafted
iancovert/sage/sage/kernel_estimator.py
pointer only (licence: MIT); the stored copy get_code serves, from nliulab/sim-shapley, is pointer only (licence: NONE) · get_code("6addeecb4c273c67")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from nliulab/sim-shapley, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

calculate_result Ran
ran; our contract draft was wrong, not the code
iancovert/sage/sage/kernel_estimator.py
pointer only (licence: MIT); the stored copy get_code serves, from nliulab/sim-shapley, is pointer only (licence: NONE) · get_code("0ad549d6b95dcb4c")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from nliulab/sim-shapley, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

completeness Ran
ran, violated the contract we drafted
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("38a0083dd2fb6fc4")
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 float32, 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
import scipy

# Create a small importance_matrix for testing
# Shape: (num_factors, num_dimensions)
# Using distinct sizes: 4 factors, 3 dimensions
importance_matrix = np.array([
    [0.5, 0.2, 0.1],
    [0.3, 0.6, 0.4],
    [0.1, 0.1, 0.3],
    [0.1, 0.1, 0.2]
], dtype=np.float32)

ARGS = {
    "importance_matrix": importance_matrix
}
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 float32, 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
completeness_per_factor Ran
ran, violated the contract we drafted
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("63c4b01b5c4d077a")
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 (3,) 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
import scipy.stats

# The function expects an importance_matrix parameter
# Looking at the body:
# - importance_matrix needs to support addition with a float (+ 1e-11)
# - importance_matrix needs to have a .shape attribute with at least shape[0]
# - It's passed to scipy.stats.entropy which expects array-like input
# - The result should be a numeric value representing completeness

# Create a small importance matrix
# Shape: (4, 3) - distinct dimensions to probe later
importance_matrix = np.array([
    [0.5, 0.3, 0.2],
    [0.1, 0.6, 0.3],
    [0.4, 0.4, 0.2],
    [0.2, 0.3, 0.5]
], dtype=np.float32)

ARGS = {
    "importance_matrix": importance_matrix
}
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 ndarray, ndim 1, dtype family float, shape (3,) — shape (3,) 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_importance_gbt Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("23d05dde2a92ea4b")
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
from sklearn import ensemble

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

# Create small synthetic data
# x_train: (num_codes, num_samples) - transposed for sklearn
# y_train: (num_factors, num_samples)
num_codes = 4
num_factors = 3
num_train_samples = 6
num_test_samples = 4

# Generate training data
x_train = np.random.randn(num_codes, num_train_samples).astype(np.float64)
y_train = np.random.randint(0, 2, size=(num_factors, num_train_samples)).astype(np.int64)

# Generate test data
x_test = np.random.randn(num_codes, num_test_samples).astype(np.float64)
y_test = np.random.randint(0, 2, size=(num_factors, num_test_samples)).astype(np.int64)

ARGS = {
    "x_train": x_train,
    "y_train": y_train,
    "x_test": x_test,
    "y_test": y_test,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
disentanglement Ran
ran, violated the contract we drafted
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("9036466b82f32c46")
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 float32, 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
import scipy

# Create a small importance matrix
# Shape: (num_codes, num_factors)
# The function uses:
# - importance_matrix.T for transpose
# - importance_matrix.shape[1] for number of codes
# - importance_matrix.sum() for total sum
# - importance_matrix.sum(axis=1) for per-code importance
importance_matrix = np.array([
    [0.8, 0.1, 0.05, 0.05],
    [0.1, 0.7, 0.1, 0.1],
    [0.05, 0.2, 0.6, 0.15],
    [0.05, 0.0, 0.25, 0.7]
], dtype=np.float32)

ARGS = {
    "importance_matrix": importance_matrix
}
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 float32, 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
disentanglement_per_code Ran
ran, violated the contract we drafted
andreinicolicioiu/dci-es/metrics/dci.py
code served (permissive licence) · get_code("4549dd9ced62ede6")
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
import scipy.stats

# Create a small importance matrix
# This represents importance scores for each code across different factors
# Shape: (n_codes, n_factors) - using 4 codes and 3 factors for distinct dimensions
importance_matrix = np.array([
    [0.8, 0.1, 0.05],
    [0.05, 0.9, 0.02],
    [0.1, 0.05, 0.85],
    [0.05, 0.05, 0.1]
], dtype=np.float32)

ARGS = {
    "importance_matrix": importance_matrix
}
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 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
estimate_holdout_importance Ran
ran on a synthesized input
iancovert/sage/sage/iterated_estimator.py
code served (permissive licence) · get_code("458e1a8bd75817bc")
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

class SimpleImputer:
    """Minimal imputer that accepts X and S (selection mask)."""
    def __init__(self, num_groups):
        self.num_groups = num_groups
    
    def __call__(self, x, S):
        # x shape: (batch_size, num_features)
        # S shape: (batch_size, num_groups) - boolean mask
        # Return imputed values of same shape as x
        return x * S.astype(np.float32)

def simple_loss_fn(y_pred, y_true):
    """Simple MSE-like loss."""
    return np.mean((y_pred - y_true) ** 2, axis=1)

# Create small test data
np.random.seed(0)
N = 10  # number of samples
num_features = 4
batch_size = 2

X = np.random.randn(N, num_features).astype(np.float32)
Y = np.random.randn(N, 1).astype(np.float32)

imputer = SimpleImputer(num_groups=num_features)

ARGS = {
    "imputer": imputer,
    "X": X,
    "Y": Y,
    "batch_size": batch_size,
    "loss_fn": simple_loss_fn,
    "batches": 3,
    "rng": np.random.RandomState(42),
}
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
estimate_total Ran
ran on a synthesized input
iancovert/sage/sage/iterated_estimator.py
code served (permissive licence) · get_code("50231e954619735e")
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 float32, 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
import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a simple imputer class that satisfies the function's usage
class SimpleImputer(nn.Module):
    def __init__(self, num_groups=4):
        super().__init__()
        self.num_groups = num_groups
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x, mask):
        # x is input data, mask is a boolean array indicating which features to use
        # Returns predictions
        x_tensor = torch.tensor(x, dtype=torch.float32)
        return self.linear(x_tensor)

# Create sample data
X = np.random.randn(6, 4).astype(np.float32)
Y = np.random.randn(6, 3).astype(np.float32)

# Create a simple loss function
def loss_fn(pred, y):
    y_tensor = torch.tensor(y, dtype=torch.float32)
    mse_loss = nn.MSELoss(reduction='none')
    return mse_loss(pred, y_tensor).mean(dim=1).detach().numpy()

# Create the imputer instance
imputer = SimpleImputer(num_groups=4)

ARGS = {
    'imputer': imputer,
    'X': X,
    'Y': Y,
    'batch_size': 2,
    'loss_fn': loss_fn
}
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 float32, 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
model_conversion Ran
ran on a synthesized input
iancovert/sage/sage/utils.py
code served (permissive licence) · get_code("3a99f611229c7477")
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 function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np

# Set seed for reproducibility
torch.manual_seed(0)

# Create a simple PyTorch model
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 2)
    
    def forward(self, x):
        return self.linear(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
numpy, torch, xgboost — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
verify_model_data Ran
ran on a synthesized input
iancovert/sage/sage/utils.py
code served (permissive licence) · get_code("1b47a6f4cbba8d7b")
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

class SimpleImputer:
    """Minimal imputer that satisfies the function's requirements."""
    def __init__(self, num_groups=3):
        self.num_groups = num_groups
    
    def __call__(self, x, mask):
        """Return probability predictions for binary classification."""
        batch_size = len(x)
        # Return probabilities for binary classification (shape: batch_size, 2)
        probs = np.random.rand(batch_size, 2)
        probs = probs / probs.sum(axis=1, keepdims=True)
        return probs

class ZeroOneLoss:
    """zero-one loss that expects probabilities."""

    def __init__(self, reduction='mean'):
        assert reduction in ('none', 'mean')
        self.reduction = reduction

    def __call__(self, pred, target):
        if pred.ndim == 1:
            pred = pred[:, np.newaxis]
        if pred.shape[1] == 1:
            pred = np.append(1 - pred, pred, axis=1)
        if target.ndim == 1:
            loss = (np.argmax(pred, axis=1) != target).astype(float)
        elif target.ndim == 2:
            loss = (np.argmax(pred, axis=1) != np.argmax(target, axis=1)).astype(float)
        else:
            raise ValueError('incorrect labels shape for zero-one loss')
        if self.reduction == 'mean':
            return np.mean(loss)
        return loss

class CrossEntropyLoss:
    """Cross entropy loss that expects probabilities."""

    def __init__(self, reduction='mean'):
        assert reduction in ('none', 'mean')
        self.reduction = reduction

    def __call__(self, pred, target, eps=1e-12):
        pred = np.clip(pred, eps, 1 - eps)
        if pred.ndim == 1:
            pred = pred[:, np.newaxis]
        if pred.shape[1] == 1:
            pred = np.append(1 - pred, pred, axis=1)
        if target.ndim == 1:
            loss = -np.log(pred[np.arange(len(pred)), target])
        elif target.ndim == 2:
            loss = -np.sum(target * np.log(pred), axis=1)
        else:
            raise ValueError('incorrect labels shape for cross entropy loss')
        if self.reduction == 'mean':
            return np.mean(loss)
        else:
            return loss

# Create sample data
imputer = SimpleImputer(num_groups=3)
X = np.random.rand(6, 4).astype(np.float32)
Y = np.array([0, 1, 0, 1, 0, 1], dtype=int)
loss = CrossEntropyLoss(reduction='mean')
batch_size = 2

ARGS = {
    'imputer': imputer,
    'X': X,
    'Y': Y,
    'loss': loss,
    'batch_size': batch_size
}
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
verify_nonoverlapping Ran
ran on a synthesized input
iancovert/sage/sage/grouped_imputers.py
code served (permissive licence) · get_code("c185110053d6cf9d")
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

# The function verify_nonoverlapping takes a parameter 'groups'
# which should be a list/sequence of array-like objects containing indices.
# It concatenates them and checks that no index appears more than once.

# Case 1: Non-overlapping groups (valid)
group1 = np.array([0, 1, 2])
group2 = np.array([3, 4, 5])
group3 = np.array([6, 7])

ARGS = {
    "groups": [group1, group2, group3]
}
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
comparison_plot Not yet run
unverified (harvested, no recorded run)
iancovert/sage/sage/plotting.py
code served (permissive licence) · get_code("854ad2bb1abd7e6f")
How it ran

no invocation recorded.

dataset_output Not yet run
unverified (harvested, no recorded run)
iancovert/sage/sage/utils.py
code served (permissive licence) · get_code("ddb117b45426e794")
How it ran

no invocation recorded.

estimate_constraints Not yet run
unverified (harvested, no recorded run)
iancovert/sage/sage/kernel_estimator.py
pointer only (licence: MIT); the stored copy get_code serves, from nliulab/sim-shapley, is pointer only (licence: NONE) · get_code("65b4135afabeaeae")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from nliulab/sim-shapley, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

plot Not yet run
unverified (harvested, no recorded run)
iancovert/sage/sage/plotting.py
code served (permissive licence) · get_code("6b00515e5abb8eda")
How it ran

no invocation recorded.

plot_sign Not yet run
unverified (harvested, no recorded run)
iancovert/sage/sage/plotting.py
code served (permissive licence) · get_code("6b966fc2c2307373")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 5 are also present in other papers whose code we hold (5 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.

FunctionAlso present inEarliest-datedOthers named
calculate_A 1 other paper
1 dated earlier
2004.00668 (2020-04)
Understanding Global Feature Contributions With Additive Importance Measures
calculate_result 1 other paper
1 dated earlier
2004.00668 (2020-04)
Understanding Global Feature Contributions With Additive Importance Measures
disentanglement
brick of this paper's assembly: metric
1 other paper
1 dated earlier
2106.03630 (2021-06)
Efficient Iterative Amortized Inference for Learning Symmetric and Disentangled Multi-Object Representations
disentanglement_per_code
brick of this paper's assembly: metric
1 other paper
1 dated earlier
2106.03630 (2021-06)
Efficient Iterative Amortized Inference for Learning Symmetric and Disentangled Multi-Object Representations
estimate_constraints 1 other paper
1 dated earlier
2004.00668 (2020-04)
Understanding Global Feature Contributions With Additive Importance Measures

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

In representation learning, a common approach is to seek representations which disentangle the underlying factors of variation. Eastwood & Williams (2018) proposed three metrics for quantifying the quality of such disentangled representations: disentanglement (D), completeness (C) and informativeness (I). In this work, we first connect this DCI framework to two common notions of linear and nonlinear identifiability, thereby establishing a formal link between disentanglement and the closely-related field of independent component analysis. We then propose an extended DCI-ES framework with two new measures of representation quality-explicitness (E) and size (S)-and point out how D and C can be computed for black-box predictors. Our main idea is that the functional capacity required to use a representation is an important but thus-far neglected aspect of representation quality, which we quantify using explicitness or ease-of-use (E). We illustrate the relevance of our extensions on the MPI3D and Cars3D datasets.

For agents

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

get_harvested_code_for_paper("2210.00364")
get_code_for_paper("2210.00364")
have("2210.00364")

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

curl https://syntology.ai/api/ran/2210.00364.json

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

[![Syntology run record](https://syntology.ai/api/ran/2210.00364.svg)](https://syntology.ai/paper/2210.00364)

Connect an agent — have() is free.