SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.18017 · NeurIPS · 2024

Transferring disentangled representations: bridging the gap between synthetic and real images

Jacopo Dapueto, Nicoletta Noceti, Francesca Odone

arXiv · PDF · Open in the Atlas

Code that ran

8 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2409.18017.json · badge: /api/ran/2409.18017.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
JacopoDapueto/transfer_disentanglement marked official in Papers with Code
Papers with Code archive
8 ran · 5 no recorded run
FunctionStatusWhere it lives
_multiple_encoding_score Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("ba48765288c4908e")
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 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

# Analyze the function:
# - dim: an integer index used to index into perfect_cov
# - idx_factor: an integer used to index into covs[:, idx_factor]
# - covs: a 2D numpy array where we access covs[:, idx_factor]

# The function creates perfect_cov with shape matching covs[:, idx_factor]
# Then sets perfect_cov[dim] = 1.0
# So covs should be 2D with shape (n_samples, n_factors)
# dim should be a valid index into the first dimension
# idx_factor should be a valid index into the second dimension

# Create small test data
n_samples = 5
n_factors = 3

# covs is a 2D array of shape (n_samples, n_factors)
covs = np.random.rand(n_samples, n_factors)

# dim should be a valid index into the first dimension (0 to n_samples-1)
dim = 2

# idx_factor should be a valid index into the second dimension (0 to n_factors-1)
idx_factor = 1

ARGS = {
    "dim": dim,
    "idx_factor": idx_factor,
    "covs": covs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
_overlapping_score Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("76d029feff99c528")
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 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 arguments for _overlapping_score
# The function takes:
# - dim: an integer index into the first dimension of covs
# - idx_factor: an integer index into the second dimension of covs
# - covs: a 2D numpy array

# Create a small 2D covariance matrix
# Using shape (4, 3) to have distinct dimensions
covs = np.random.rand(4, 3)

# dim should be a valid index into the first dimension (0-3)
dim = 1

# idx_factor should be a valid index into the second dimension (0-2)
idx_factor = 2

ARGS = {
    "dim": dim,
    "idx_factor": idx_factor,
    "covs": covs
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
associate_dims_factors Ran
ran; our contract draft was wrong, not the code
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("1a06edc1a3d86117")
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

# Read the function to understand what it needs:
# - covs: a 2D numpy array, indexed as covs[dim, :] and covs.shape[0]
# - factors: an iterable that can be enumerated, len(factors) determines n_factors

# Create a small covs matrix (n_features x n_factors)
# n_features = 4 (number of dimensions)
# n_factors = 3 (number of factors)
covs = np.array([
    [0.9, 0.05, 0.05],
    [0.1, 0.8, 0.1],
    [0.05, 0.1, 0.85],
    [0.2, 0.3, 0.5]
], dtype=np.float32)

# Create a small factors list/array
factors = ['factor_0', 'factor_1', 'factor_2']

ARGS = {
    'covs': covs,
    'factors': factors
}
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
disentanglement_score Ran
ran; our contract draft was wrong, not the code
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("19ff3fdd8ecdf5d1")
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 tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small covariance matrix
# Shape: (num_dims, num_factors)
# num_dims = 4, num_factors = 3
covs = np.array([
    [0.9, 0.1, 0.0],
    [0.1, 0.8, 0.1],
    [0.0, 0.1, 0.9],
    [0.0, 0.0, 0.1]
], dtype=np.float32)

# List of factor names
factors = ['factor_0', 'factor_1', 'factor_2']

# Alpha parameter for weighting overlap vs encoding
alpha = 0.5

# Gamma parameter for aggregation method
gamma = 'avg'

ARGS = {
    'covs': covs,
    'factors': factors,
    'alpha': alpha,
    'gamma': gamma
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, 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
is_inactive Ran
ran; our contract draft was wrong, not the code
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("b7756c2e00377ac0")
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 ndarray, ndim 1, dtype family float, shape (8,) 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 representation matrix
# The function accesses representation.shape[0] and uses indexing with [eval_representation_pos, :]
# It computes variance along axis 0, so we need a 2D array
np.random.seed(0)
representation = np.random.randn(100, 8)

ARGS = {
    "representation": representation,
    "n_eval": 10000,
    "threshold": 0.05,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 (8,) — shape (8,) 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
multiple_encoding_score Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("70547cf6c6866d51")
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 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 arguments for multiple_encoding_score function
# The function expects:
# - idx_factor: an integer index
# - covs: a 2D numpy array where covs[:, idx_factor] is accessed
# - gamma: either 'avg' or something else (used in conditional)

# Create a small covariance matrix
# Shape: (4, 3) - 4 dimensions, 3 factors
# This ensures covs.shape[0] = 4 for the loop in multiple_encoding_score
covs = np.array([
    [0.9, 0.1, 0.2],
    [0.1, 0.8, 0.3],
    [0.05, 0.05, 0.7],
    [0.05, 0.05, 0.0]
], dtype=np.float32)

# idx_factor should be a valid column index (0, 1, or 2)
idx_factor = 1

# gamma can be 'avg' or any other value
gamma = 'avg'

ARGS = {
    'idx_factor': idx_factor,
    'covs': covs,
    'gamma': gamma
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
overlapping_score Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("107b7fca0c86be71")
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 overlapping_score function

# Create a small coverage matrix
# Shape: (n_dims, n_factors)
# The function iterates over dims (rows) and accesses covs[:, idx_factor]
covs = np.array([
    [0.9, 0.1, 0.0],
    [0.2, 0.7, 0.1],
    [0.05, 0.05, 0.9],
    [0.3, 0.3, 0.4],
], dtype=np.float32)

# idx_factor should be a valid column index in covs
idx_factor = 1

# gamma can be False or 'avg'
gamma = 'avg'

ARGS = {
    'idx_factor': idx_factor,
    'covs': covs,
    'gamma': gamma,
}
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
random_samples Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("36241753b06b1961")
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 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

# Create sample data
np.random.seed(0)
x = np.random.randn(1000, 4).astype(np.float32)
y = np.random.randint(0, 2, size=(1000,))

ARGS = {
    "x": x,
    "y": y,
    "n": 100
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
Metric Not yet run
unverified (harvested, no recorded run)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("e15d4aad44911318")
How it ran

no invocation recorded.

OMES Not yet run
unverified (harvested, no recorded run)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("5bbb51c9da670e3c")
How it ran

no invocation recorded.

cov_factor Not yet run
unverified (harvested, no recorded run)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("1fffa78eb30c0230")
How it ran

no invocation recorded.

get_score Not yet run
unverified (harvested, no recorded run)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("bbe86af17ad979c9")
How it ran

no invocation recorded.

plot_cov_factors Not yet run
unverified (harvested, no recorded run)
JacopoDapueto/transfer_disentanglement/src/evaluation/metrics/omes.py
code served (permissive licence) · get_code("6e12cb2b85438ed6")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 13 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Developing meaningful and efficient representations that separate the fundamental structure of the data generation mechanism is crucial in representation learning. However, Disentangled Representation Learning has not fully shown its potential on real images, because of correlated generative factors, their resolution and limited access to ground truth labels. Specifically on the latter, we investigate the possibility of leveraging synthetic data to learn general-purpose disentangled representations applicable to real data, discussing the effect of fine-tuning and what properties of disentanglement are preserved after the transfer. We provide an extensive empirical study to address these issues. In addition, we propose a new interpretable intervention-based metric, to measure the quality of factors encoding in the representation. Our results indicate that some level of disentanglement, transferring a representation from synthetic to real data, is possible and effective.

For agents

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

get_harvested_code_for_paper("2409.18017")
get_code_for_paper("2409.18017")
have("2409.18017")

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

curl https://syntology.ai/api/ran/2409.18017.json

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

[![Syntology run record](https://syntology.ai/api/ran/2409.18017.svg)](https://syntology.ai/paper/2409.18017)

Connect an agent — have() is free.