SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.07668 · ICCV · 2021

Orthogonal Jacobian Regularization for Unsupervised Disentanglement in Image Generation

Yuan Gao, Wangmeng Zuo, Xiao Liu, Yuxiang Wei, Zhilong Ji, Zhongqin Wu, Yupeng Shi

arXiv · PDF · Open in the Atlas

Code that ran

20 of 21 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, 0 violated it and 6 have no contract verdict; 13 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2108.07668.json · badge: /api/ran/2108.07668.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
csyxwei/orojar canonical (classifier label)
classifier basis: census
16 ran · 1 no recorded run
csyxwei/OroJaR community
Papers with Code archive
4 ran · 0 no recorded run
FunctionStatusWhere it lives
D_arch Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/BigGAN.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from VICO-UoE/IT-GAN, is pointer only (licence: NONE) · get_code("85eb95cce203dfc7")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from VICO-UoE/IT-GAN, 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.

D_arch Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/BigGANdeep.py
code served (permissive licence) · get_code("a99a05ac52533be9")
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 torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    'ch': 64,
    'attention': '64',
    'ksize': '333333',
    'dilation': '111111'
}
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 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
G_arch Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/BigGAN.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from VICO-UoE/IT-GAN, is pointer only (licence: NONE) · get_code("e8c308c9283407d7")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from VICO-UoE/IT-GAN, 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.

G_arch Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/BigGANdeep.py
code served (permissive licence) · get_code("44d6a6343f6eb182")
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 torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    'ch': 64,
    'attention': '64_128_256',
    'ksize': '333333',
    'dilation': '111111'
}
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 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
build_job_name Ran
ran on a synthesized input
csyxwei/orojar/progan_experiments/train_orojar.py
code served (permissive licence) · get_code("4da4d6bc0bd3244a")
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 str. 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.

class Options:
    def __init__(self):
        self.hp_lambda = 0
        self.infogan_lambda = 0
        self.dataset = "mnist"
        self.nz = 100
        self.infogan_nz = 10

ARGS = {
    'opt': Options()
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
first_directional_derivative Ran
ran on a synthesized input
csyxwei/orojar/biggan_discovery/orojar.py
code served (permissive licence) · get_code("09e8b81f44dc0143")
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, 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 torch.nn as nn

torch.manual_seed(0)

# Define a simple generator model that takes z, c, optional w, and optional Q
class SimpleGenerator(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, z, c, w=None, Q=None):
        # z: input tensor
        # c: conditioning tensor
        # w: optional weight/style tensor
        # Q: optional query tensor
        if w is not None:
            z = z + w
        output = self.linear(z)
        if c is not None:
            output = output + c
        if Q is not None:
            output = output + Q
        return output

# Instantiate the generator
G = SimpleGenerator()

# Create small tensors
z = torch.randn(2, 4)  # batch size 2, dim 4
c = torch.randn(2, 3)  # conditioning, compatible with output
x = torch.randn(2, 4)  # direction for z, same shape as z
epsilon = 1e-4

# Compute G_z for the current z
G_z = G(z, c)

# Optional parameters
w = torch.randn(2, 4)  # optional weight tensor, same shape as z
Q = torch.randn(2, 3)  # optional query tensor, compatible with output

ARGS = {
    'G': G,
    'z': z,
    'c': c,
    'x': x,
    'G_z': G_z,
    'epsilon': epsilon,
    'w': w,
    'Q': Q
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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, 3) — shape (2, 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
get_current_hessian_penalty_loss_weight Ran
ran on a synthesized input
csyxwei/orojar/progan_experiments/training/orojar.py
code served (permissive licence) · get_code("84a7e309cd2bf8f6")
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 float, 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 torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "max_lambda": 0.1,
    "hp_start_iter": 100,
    "t": 150,
    "T": 50
}
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 float, 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
gram_schmidt Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/layers.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from mshahbazi72/cGANTransfer, is pointer only (licence: NONE) · get_code("c41836f3e51fa8f1")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from mshahbazi72/cGANTransfer, 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.

listify Ran
ran, honoured the contract we drafted
csyxwei/orojar/biggan_discovery/orojar.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from zhihengli-UR/discover_unknown_biases, is pointer only (licence: NONE) · get_code("47b42795fd222ceb")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from zhihengli-UR/discover_unknown_biases, 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.

loss_dcgan_dis Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/losses.py
code served (permissive licence) · get_code("982a26492dc6b7fb")
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 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create small discriminator output tensors
# dis_fake: discriminator output for fake samples
dis_fake = torch.randn(2, 4)

# dis_real: discriminator output for real samples
dis_real = torch.randn(2, 4)

ARGS = {
    "dis_fake": dis_fake,
    "dis_real": dis_real,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
loss_dcgan_gen Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/losses.py
code served (permissive licence) · get_code("6c34831223553b6f")
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 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# The function loss_dcgan_gen expects dis_fake, which is used with:
# - F.softplus() - expects a tensor
# - torch.mean() - expects a tensor
# So dis_fake should be a tensor

# Create a small fake discriminator output tensor (batch_size=2, some features)
dis_fake = torch.randn(2, 4)

ARGS = {
    "dis_fake": dis_fake,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
loss_hinge_dis Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/biggan_discovery/losses.py
code served (permissive licence) · get_code("290cd37ab0098d5c")
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 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# The function loss_hinge_dis takes two tensor arguments:
# dis_fake and dis_real
# It computes relu-based hinge losses on them and returns a tuple of two scalars

# Create small discriminator output tensors (batch_size=2, arbitrary values)
dis_fake = torch.randn(2, 4)  # Fake discriminator scores
dis_real = torch.randn(2, 4)  # Real discriminator scores

ARGS = {
    "dis_fake": dis_fake,
    "dis_real": dis_real,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
multi_layer_first_directional_derivative Ran
ran; our contract draft was wrong, not the code
csyxwei/OroJaR/orojar_pytorch.py
code served (permissive licence) · get_code("f6b513df7a8f90bf")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple neural network module for G
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create model instance
G = SimpleModel()

# Create input tensors with distinct sizes
z = torch.randn(2, 4)  # batch size 2, dimension 4
x = torch.randn(2, 4)  # direction vector, same shape as z
G_z = G(z)  # output of G at z

# Small epsilon for numerical differentiation
epsilon = 1e-4

# Empty kwargs (no additional keyword arguments needed for this simple model)
G_kwargs = {}

ARGS = {
    'G': G,
    'z': z,
    'x': x,
    'G_z': G_z,
    'epsilon': epsilon,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
multi_stack_var_and_reduce Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
csyxwei/OroJaR/orojar_pytorch.py
code served (permissive licence) · get_code("16fa0b0146a66a69")
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 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 torch

torch.manual_seed(0)

# The function multi_stack_var_and_reduce takes:
# - fdds: an iterable of iterables of activations (tensors)
# - reduction: a reduction function (default torch.max)
# - return_separately: a boolean (default False)

# Looking at the function body:
# - It zips(*fdds), so fdds should be a list/tuple of lists/tuples of tensors
# - Each activation tensor is used in stack_var_and_reduce
# - stack_var_and_reduce does:
#   - x.view(x.size(0), 1, -1) and x.view(x.size(0), -1, 1)
#   - torch.bmm on these reshaped tensors
#   - torch.stack on the results
#   - torch.var with dim=0
#   - reduction(var_tensor)

# Create small activation tensors: batch_size=2, feature_dim=4
# We'll create multiple activation lists to pass to fdds

# First set of activations (3 tensors)
activ_set_1 = [
    torch.randn(2, 4),
    torch.randn(2, 4),
    torch.randn(2, 4),
]

# Second set of activations (3 tensors, same structure for zipping)
activ_set_2 = [
    torch.randn(2, 4),
    torch.randn(2, 4),
    torch.randn(2, 4),
]

# Third set of activations (3 tensors)
activ_set_3 = [
    torch.randn(2, 4),
    torch.randn(2, 4),
    torch.randn(2, 4),
]

# fdds should be a list where each element is an iterable of tensors
# When we zip(*fdds), we get tuples of tensors that correspond across the different activation sets
fdds = [activ_set_1, activ_set_2, activ_set_3]

ARGS = {
    "fdds": fdds,
    "reduction": torch.max,
    "return_separately": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
norm Ran
ran on a synthesized input
csyxwei/orojar/biggan_discovery/layers.py
code served (permissive licence) · get_code("2de3863ec32c1c1f")
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

torch.manual_seed(0)

# The function norm() takes a tensor rr and normalizes it along dimension 1
# It computes the norm along dim=1 with keepdim=True, then divides rr by that norm
# We need a 2D tensor (batch_size, features)

rr = torch.randn(2, 4)

ARGS = {
    "rr": rr
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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
orojar Ran
ran; our contract draft was wrong, not the code
csyxwei/OroJaR/orojar_pytorch.py
code served (permissive licence) · get_code("a2ed9ad66857123a")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple neural network module that returns a list of activations
class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 5)
        self.linear2 = nn.Linear(5, 3)
    
    def forward(self, x, **kwargs):
        # Return a list of activations (intermediate and final)
        h1 = self.linear1(x)
        h2 = self.linear2(h1)
        return [h1, h2]

# Create an instance of the network
G = SimpleNet()

# Create input tensor z with shape (batch_size=2, input_dim=4)
z = torch.randn(2, 4)

# Define G_kwargs (can be empty or contain additional arguments like class labels)
G_kwargs = {}

# Pre-compute G_z to test the optional parameter
G_z = G(z, **G_kwargs)

# Build ARGS dictionary with all parameters for orojar
ARGS = {
    'G': G,
    'z': z,
    'k': 2,
    'epsilon': 0.1,
    'reduction': torch.max,
    'return_separately': False,
    'G_z': G_z,
    'G_kwargs': G_kwargs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
orojar Ran
ran on a synthesized input
csyxwei/orojar/orojar_pytorch.py
code served (permissive licence) · get_code("ff29a0906a2963fb")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple neural network that returns activations
class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(4, 3)
        self.fc2 = nn.Linear(3, 2)
    
    def forward(self, z, **kwargs):
        # Return list of activations as the function expects
        x = torch.relu(self.fc1(z))
        y = self.fc2(x)
        return [x, y]

# Create the network
G = SimpleNet()

# Create input tensor: batch_size=2, input_dim=4
z = torch.randn(2, 4)

# Compute G(z) once to pass as G_z (optional optimization)
G_z = G(z)

# Define other parameters
k = 2  # number of Jacobian directions
epsilon = 0.1  # blur amount
reduction = torch.max  # reduction function

# Build the arguments dictionary
ARGS = {
    'G': G,
    'z': z,
    'k': k,
    'epsilon': epsilon,
    'reduction': reduction,
    'return_separately': False,
    'G_z': G_z,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
rademacher Ran
ran; our contract draft was wrong, not the code
csyxwei/orojar/orojar_pytorch.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from zhihengli-UR/discover_unknown_biases, is pointer only (licence: NONE) · get_code("8de600fe2220e241")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from zhihengli-UR/discover_unknown_biases, 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.

rademacher Ran
ran on a synthesized input
csyxwei/orojar/biggan_discovery/orojar.py
code served (permissive licence) · get_code("58800b4db4b14ffd")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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

torch.manual_seed(0)

ARGS = {
    "shape": (2, 4),
    "gpu": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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
stack_var_and_reduce Ran
ran; our contract draft was wrong, not the code
csyxwei/OroJaR/orojar_pytorch.py
code served (permissive licence) · get_code("97ecfd58b1dab254")
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 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 torch

torch.manual_seed(0)

# The function expects:
# 1. list_of_activations: a list of tensors
# 2. reduction: a reduction function (default torch.max)

# Looking at the function body:
# - Each x in list_of_activations is used with x.view() and bmm operations
# - x needs to support .size() and .view()
# - The tensors are reshaped to (batch_size, 1, -1) and (batch_size, -1, 1)
# - torch.bmm expects 3D tensors of shape (batch, n, m) and (batch, m, p)
# - Result is stacked and var is computed

# Create a list of small activation tensors
# Shape: (batch_size=2, features=4)
activation_1 = torch.randn(2, 4)
activation_2 = torch.randn(2, 4)
activation_3 = torch.randn(2, 4)

list_of_activations = [activation_1, activation_2, activation_3]

# reduction function - use torch.max as default
reduction_fn = torch.max

ARGS = {
    "list_of_activations": list_of_activations,
    "reduction": reduction_fn,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
proj Not yet run
unverified (harvested, no recorded run)
csyxwei/orojar/biggan_discovery/layers.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from mshahbazi72/cGANTransfer, is pointer only (licence: NONE) · get_code("4015ab1f3ab9b881")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from mshahbazi72/cGANTransfer, 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.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 13 are also present in other papers whose code we hold (13 with at least one paper dated earlier, 1 the same month, 9 later) and 7 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
gram_schmidt 28 other papers
12 dated earlier, 15 later, 1 same year, order not determinable · 5 only as an assembly brick
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1910.05852 (2019-10)
Implicit competitive regularization in GANs
and 25 more not named here
proj 28 other papers
12 dated earlier, 15 later, 1 same year, order not determinable · 5 only as an assembly brick
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1910.05852 (2019-10)
Implicit competitive regularization in GANs
and 25 more not named here
G_arch 18 other papers
7 dated earlier, 1 the same month, 10 later · 3 only as an assembly brick
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
and 15 more not named here
D_arch 15 other papers
6 dated earlier, 9 later
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
and 12 more not named here
loss_hinge_dis 11 other papers
4 dated earlier, 7 later
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1912.00953 (2019-12)
LOGAN: Latent Optimisation for Generative Adversarial Networks
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
and 8 more not named here
loss_dcgan_dis 10 other papers
3 dated earlier, 7 later
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
and 7 more not named here
G_arch 9 other papers
2 dated earlier, 7 later
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
aaai_20247 (2022 (year only))
An Adversarial Framework for Generating Unseen Images by Activation Maximization
and 6 more not named here
loss_dcgan_gen 9 other papers
2 dated earlier, 7 later
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
aaai_20247 (2022 (year only))
An Adversarial Framework for Generating Unseen Images by Activation Maximization
and 6 more not named here
D_arch 8 other papers
2 dated earlier, 6 later
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
aaai_20247 (2022 (year only))
An Adversarial Framework for Generating Unseen Images by Activation Maximization
and 5 more not named here
rademacher
brick of this paper's assembly: helper
2 other papers
2 dated earlier
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
2104.14556 (2021-04)
Discover the Unknown Biased Attribute of an Image Classifier
get_current_hessian_penalty_loss_weight 1 other paper
1 dated earlier
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
listify
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2104.14556 (2021-04)
Discover the Unknown Biased Attribute of an Image Classifier
rademacher 1 other paper
1 dated earlier
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement

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

Unsupervised disentanglement learning is a crucial issue for understanding and exploiting deep generative models. Recently, SeFa tries to find latent disentangled directions by performing SVD on the first projection of a pretrained GAN. However, it is only applied to the first layer and works in a post-processing way. Hessian Penalty minimizes the off-diagonal entries of the output's Hessian matrix to facilitate disentanglement, and can be applied to multi-layers. However, it constrains each entry of output independently, making it not sufficient in disentangling the latent directions (e.g., shape, size, rotation, etc.) of spatially correlated variations. In this paper, we propose a simple Orthogonal Jacobian Regularization (OroJaR) to encourage deep generative model to learn disentangled representations. It simply encourages the variation of output caused by perturbations on different latent dimensions to be orthogonal, and the Jacobian with respect to the input is calculated to represent this variation. We show that our OroJaR also encourages the output's Hessian matrix to be diagonal in an indirect manner. In contrast to the Hessian Penalty, our OroJaR constrains the output in a holistic way, making it very effective in disentangling latent dimensions corresponding to spatially correlated variations. Quantitative and qualitative experimental results show that our method is effective in disentangled and controllable image generation, and performs favorably against the state-of-the-art methods. Our code is available at https://github.com/csyxwei/OroJaR.

For agents

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

get_harvested_code_for_paper("2108.07668")
get_code_for_paper("2108.07668")
have("2108.07668")

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

curl https://syntology.ai/api/ran/2108.07668.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.07668.svg)](https://syntology.ai/paper/2108.07668)

Connect an agent — have() is free.