SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.15798 · NeurIPS · 2021

Rethinking Neural Operations for Diverse Tasks

Christopher Ré, Ameet Talwalkar, Liam Li, Mikhail Khodak, Nicholas Roberts, Tri Dao

arXiv · PDF · Open in the Atlas

Code that ran

22 of 34 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (8 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 9 failures of our instrument; 12 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/2103.15798.json · badge: /api/ran/2103.15798.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
mkhodak/relax canonical (classifier label)
classifier basis: classifier_2025
15 ran · 9 no recorded run
nick11roberts/XD canonical (classifier label)
classifier basis: census
4 ran · 2 no recorded run
nick11roberts/xd canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
ba-lab/pdnet application (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
AvgPool Ran
ran; our contract draft was wrong, not the code
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("2a31a6dcb06ddf63")
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 type. 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
from torch import nn

torch.manual_seed(0)

ARGS = {
    'dims': 2
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
Complex2Real Ran
ran on a synthesized input
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("bec5cac219d82ded")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) 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
from torch import nn

class Complex2Real(nn.Module):
    """module that returns real part of a complex input"""

    def forward(self, x):
        return x.real

INIT_ARGS = {}

ARGS = {
    'x': torch.complex(torch.randn(2, 4), torch.randn(2, 4))
}

if __name__ == '__main__':
    module = Complex2Real(**INIT_ARGS)
    output = module(**ARGS)
    print(output)
    print(output.shape)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
ConvPerm Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("ba89e07cd33d1453")
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 an instance of its own class, ConvPerm: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn
from functools import lru_cache

torch.manual_seed(0)

@lru_cache(maxsize=None)
def atrous_permutation(n, k, d):
    """computes single-dimensional atrous permutation for dilating convolutions
    Args:
        n: input size
        k: kernel size
        d: dilation rate
    Returns:
        permutation as a torch.LongTensor
    """
    perm = torch.arange(n).roll(k // 2 - k).flip(0)
    for i in range(k - 1, 0, -1):
        perm[i], perm[d * i] = (perm[d * i].item(), perm[i].item())
    perm[:d * (k - 1) + 1] = perm[:d * (k - 1) + 1].flip(0)
    return perm.roll(-((d * (k - 1) + 1) // 2))

class ConvPerm(nn.Module):
    """computes permutation for convolving via FFT"""

    def __init__(self, kernel_size=1, dilation=1):
        """
        Args:
            kernel_size: kernel size of target convolution
            dilation: dilation rate of target convolution
        """
        super(ConvPerm, self).__init__()
        self.kernel_size = kernel_size
        self.dilation = dilation

    def forward(self, x):
        x = x.flip(-1).roll((self.kernel_size + 1) // 2, dims=-1)
        if self.dilation > 1:
            return x[..., atrous_permutation(x.shape[-1], self.kernel_size, self.dilation)]
        return x

# Create a test input tensor: batch_size=2, feature_dim=4, sequence_length=16
test_input = torch.randn(2, 4, 16)

ARGS = {
    "kernel_size": 3,
    "dilation": 2
}
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 ConvPerm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a ConvPerm 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
FNOPerm Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("746f20753fb4022f")
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 an instance of its own class, FNOPerm: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn

torch.manual_seed(0)

class FNOPerm(nn.Module):
    """computes permutation for applying FNO"""

    def __init__(self, kernel_size=1):
        """
        Args:
            kernel_size: kernel size
        """
        super(FNOPerm, self).__init__()
        self.kernel_size = kernel_size

    def forward(self, x):
        k = self.kernel_size
        return torch.cat([x[..., :k // 2], x[..., k:], x[..., k // 2:k]], dim=-1)


# Create a small test tensor with shape (batch=2, feature_dim=4)
x = torch.randn(2, 4)

ARGS = {
    "kernel_size": 2,
}
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 FNOPerm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a FNOPerm 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
Fourier Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("4b400ff00a487d94")
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 an instance of its own class, Fourier: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn

torch.manual_seed(0)

ARGS = {
    "inv": False,
    "normalized": True,
    "dims": 2,
    "compact": False,
    "odd": 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 Fourier — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Fourier 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
MaxPool Ran
ran; our contract draft was wrong, not the code
nick11roberts/XD/chrysalis/darts.py
code served (permissive licence) · get_code("1ff4c96878cdae81")
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 type. 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
from torch import nn

torch.manual_seed(0)

# The function MaxPool(dims) returns a MaxPool class from nn module
# based on the dims parameter. It constructs the class name dynamically
# by concatenating 'MaxPool' + str(dims) + 'd'
# For example: MaxPool(2) returns nn.MaxPool2d, MaxPool(3) returns nn.MaxPool3d

# The function takes a single parameter: dims (an integer)
# Valid values are typically 1, 2, or 3 for MaxPool1d, MaxPool2d, MaxPool3d

ARGS = {
    'dims': 2
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
Pad Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("1dde65deadfc1f2d")
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 an instance of its own class, Pad: the run constructed the object and did not call it, so no result was computed. 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 math
from torch import nn

torch.manual_seed(0)

ARGS = {
    'dims': 2,
    'in_size': (8, 8),
    'kernel_size': 3,
    'dilation': 1,
    'padding': 1,
    'padding_mode': 'circular',
    'power': 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 Pad — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Pad 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
TransposeFlip Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("7113106c9caa90dc")
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 an instance of its own class, TransposeFlip: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn

torch.manual_seed(0)

class TransposeFlip(nn.Module):
    """flips weights for transposed convolving via FFT"""

    def __init__(self, kernel_size=1):
        """
        Args:
            kernel_size: kernel size of target transposed convolution
        """
        super(TransposeFlip, self).__init__()
        self.kernel_size = kernel_size

    def forward(self, x):
        return torch.cat([x[..., :self.kernel_size].flip(-1), x[..., self.kernel_size:]], dim=-1)

# Create a small input tensor: shape (2, 4, 8) where batch=2, channels=4, length=8
x = torch.randn(2, 4, 8)

# kernel_size should be less than the last dimension of x to demonstrate the split behavior
kernel_size = 3

ARGS = {
    "kernel_size": kernel_size
}
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 TransposeFlip — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransposeFlip 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
TransposePad Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("63dc4a7a7bfdd3ba")
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 an instance of its own class, TransposePad: the run constructed the object and did not call it, so no result was computed. 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)

# Create a small input tensor for testing
# Shape: (batch=2, channels=4, length=8)
x = torch.randn(2, 4, 8)

# Define the arguments for TransposePad
ARGS = {
    'kernel_size': 3,
    'dilation': 2
}
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 TransposePad — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransposePad 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
TransposePerm Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("6817170fe598f430")
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 an instance of its own class, TransposePerm: the run constructed the object and did not call it, so no result was computed. 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
from torch import nn

torch.manual_seed(0)

ARGS = {
    "kernel_size": 3,
    "dilation": 1,
}
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 TransposePerm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransposePerm 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
TruncateHermitian Ran
ran on a synthesized input
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("df99f62038dc6b95")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import nn

class TruncateHermitian(nn.Module):
    """applies Hermitian truncation to last dimension of the Fourier transform of a real signal"""

    def forward(self, x):
        return x[..., :x.shape[-1] // 2 + 1]


INIT_ARGS = {}

ARGS = {
    'x': torch.randn(2, 4, 16)
}

if __name__ == '__main__':
    module = TruncateHermitian(**INIT_ARGS)
    output = module(**ARGS)
    print(f"Input shape: {ARGS['x'].shape}")
    print(f"Output shape: {output.shape}")
    print(f"Output:\n{output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 3, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
Unpad Ran
ran on a synthesized input · constructed an object; no result was computed
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("83fd4dada951a4ce")
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 an instance of its own class, Unpad: the run constructed the object and did not call it, so no result was computed. 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)

# Test case 1: 2D case with stride=1 (no subsampling)
dims_2d = 2
stride_2d = 1

unpad_module_2d = nn.Module()

# Create a small test tensor: batch=2, channels=3, height=4, width=5
x_2d = torch.randn(2, 3, 4, 5)

# Create an unpad tuple (boolean indexing array)
unpad_2d = (slice(None), slice(None), slice(0, 3), slice(0, 4))

ARGS = {
    "dims": dims_2d,
    "stride": stride_2d,
}
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 Unpad — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Unpad 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
accuracy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
nick11roberts/XD/DARTS/utils.py
code served (permissive licence) · get_code("e93d9ddbfeb73ced")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 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

torch.manual_seed(0)

# Create small tensors for the accuracy function
batch_size = 2
num_classes = 4

# output: logits/scores from model, shape (batch_size, num_classes)
output = torch.randn(batch_size, num_classes)

# target: ground truth labels, shape (batch_size,)
target = torch.tensor([0, 2])

# topk: tuple of k values to compute accuracy for
topk = (1, 2)

ARGS = {
    "output": output,
    "target": target,
    "topk": topk
}
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, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
atrous_permutation Ran
ran, honoured the contract we drafted
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("8597cc597b0fd17f")
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 Tensor, ndim 1, dtype family int, shape (16,) 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)

ARGS = {
    "n": 16,
    "k": 3,
    "d": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, dtype family int, shape (16,) — shape (16,) 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
check_weight_norm Ran
ran; our contract draft was wrong, not the code
nick11roberts/XD/chrysalis/chrysalis.py
code served (permissive licence) · get_code("2480ed0be135e57a")
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 WeightNorm. 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)

# Create a module with weight norm applied
module = nn.Conv1d(1, 1, 1)
torch.nn.utils.weight_norm(module)

ARGS = {
    "module": module
}
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 WeightNorm — a WeightNorm output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a WeightNorm 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
complex_dist Ran
ran; our contract draft was wrong, not the code
nick11roberts/xd/chrysalis/xd.py
code served (permissive licence) · get_code("9ec5d46f798ca92d")
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 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)

# Create small complex tensors
X = torch.randn(2, 4, dtype=torch.complex64)
Y = torch.randn(2, 4, dtype=torch.complex64)

# Define ARGS dict with all parameters for complex_dist
ARGS = {
    "X": X,
    "Y": Y,
    "relative": False,
    "scaled": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
complex_norm Ran
ran; our contract draft was wrong, not the code
nick11roberts/xd/chrysalis/xd.py
code served (permissive licence) · get_code("f99717b5f5887edd")
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 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)

# Create a small complex tensor
# The function accesses .real, .imag, .square(), and .sum()
# So we need a complex-valued tensor
X = torch.complex(
    torch.randn(2, 4),
    torch.randn(2, 4)
)

ARGS = {
    "X": X
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
get_module Ran
ran; our contract draft was wrong, not the code
nick11roberts/XD/chrysalis/chrysalis.py
code served (permissive licence) · get_code("e151df3b5424a3fb")
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 Linear. 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)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1 = nn.Linear(4, 3)
        self.layer2 = nn.Sequential(
            nn.Linear(3, 2),
            nn.ReLU()
        )
        self.submodule = nn.Module()
        self.submodule.nested = nn.Linear(2, 1)
    
    def forward(self, x):
        return self.layer1(x)

model = SimpleModel()

ARGS = {
    "model": model,
    "module_string": "layer2.0"
}
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 Linear — a Linear output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Linear 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
int2tuple Ran
ran, honoured the contract we drafted
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("b65e5411454bdb5b")
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 int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "int_or_tuple": 5,
    "length": 2,
    "allow_other": False,
}
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 tuple, ndim 1, dtype family int — 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
special_dist Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
nick11roberts/xd/chrysalis/xd.py
code served (permissive licence) · get_code("cc788ee6ee06b4df")
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 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)

# Create two small tensors for X and Y
X = torch.randn(2, 4)
Y = torch.randn(2, 4)

# Define ARGS dict with all parameters
ARGS = {
    "X": X,
    "Y": Y,
    "relative": False,
    "scaled": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
transpose_permutation Ran
ran, honoured the contract we drafted
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("c4904be3964ec72b")
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 Tensor, ndim 1, dtype family int, shape (5,) 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)

ARGS = {
    "n": 5,
    "k": 3,
    "d": 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, dtype family int, shape (5,) — shape (5,) 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
truncate_freq Ran
ran; our contract draft was wrong, not the code
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("7378c3d22e68d903")
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 generator. 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
from itertools import product

torch.manual_seed(0)

# Test case 1: 1D input
in_size_1d = [8]
freqs_1d = [3]

# Test case 2: 2D input
in_size_2d = [8, 6]
freqs_2d = [3, 2]

# Test case 3: 3D input
in_size_3d = [8, 6, 4]
freqs_3d = [3, 2, 1]

# For the function to work, we need:
# - in_size: a sequence of integers (list or tuple)
# - freqs: a sequence of integers (list or tuple) same length as in_size

# Testing with 2D case as primary example
ARGS = {
    "in_size": [8, 6],
    "freqs": [3, 2]
}
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 generator — a generator output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a generator 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
SharedOperation Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("94357aa818c939d9")
How it ran

no invocation recorded.

TensorProduct Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("da15dcc8176fc0a6")
How it ran

no invocation recorded.

XD Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("0a723f7b3a1f7560")
How it ran

no invocation recorded.

basic_fcn Not yet run
unverified (harvested, no recorded run)
ba-lab/pdnet/models.py
code served (permissive licence) · get_code("68a0988ac016a394")
How it ran

no invocation recorded.

count_parameters_in_MB Not yet run
unverified (harvested, no recorded run)
nick11roberts/XD/DARTS/utils.py
code served (permissive licence) · get_code("3e816289579cc9e9")
How it ran

no invocation recorded.

drop_path Not yet run
unverified (harvested, no recorded run)
nick11roberts/XD/DARTS/utils.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from hyeonjeongha/croze, is pointer only (licence: NONE) · get_code("13c2719404e25b25")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from hyeonjeongha/croze, 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.

fixed Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("50254999b8752976")
How it ran

no invocation recorded.

multichannel_prod Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("50200b49d8c95110")
How it ran

no invocation recorded.

pad2size Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/relax/xd.py
code served (permissive licence) · get_code("bc0b1e3f81223bd5")
How it ran

no invocation recorded.

resnet20 Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/examples/resnet/resnet.py
code served (permissive licence) · get_code("b3920efd145667ac")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

resnet32 Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/examples/resnet/resnet.py
code served (permissive licence) · get_code("ccbabd3d954fa0a8")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

resnet44 Not yet run
unverified (harvested, no recorded run)
mkhodak/relax/examples/resnet/resnet.py
code served (permissive licence) · get_code("595fe4a7995fd021")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

Also present in other papers' code

Of 34 samples lifted from this paper's linked repositories, 34 carry a function identity and 0 do not. Of the 34 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 4 later) and 30 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
count_parameters_in_MB 8 other papers
7 dated earlier, 1 later
1806.09055 (2018-06)
DARTS: Differentiable Architecture Search
1912.00195 (2019-12)
SGAS: Sequential Greedy Architecture Search
2004.14557 (2020-04)
Learning Deformable Image Registration from Optimization: Perspective, Modules, Bilevel Training and Beyond
and 5 more not named here
accuracy 7 other papers
1 dated earlier, 6 later
1806.09055 (2018-06)
DARTS: Differentiable Architecture Search
2109.03508 (2021-09)
RepNAS: Searching for Efficient Re-parameterizing Blocks
aaai_26076 (2023 (year only))
PINAT: A Permutation INvariance Augmented Transformer for NAS Predictor
and 4 more not named here
drop_path 6 other papers
5 dated earlier, 1 later
1806.09055 (2018-06)
DARTS: Differentiable Architecture Search
1912.00195 (2019-12)
SGAS: Sequential Greedy Architecture Search
2006.16537 (2020-06)
Theory-Inspired Path-Regularized Differential Network Architecture Search
and 3 more not named here
resnet20 2 other papers
2 later
2206.08464 (2022-06)
PRANC: Pseudo RAndom Networks for Compacting deep models
2408.01978 (2024-08)
AdvQDet: Detecting Query-Based Adversarial Attacks with Adversarial Contrastive Prompt Tuning

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

An important goal of AutoML is to automate-away the design of neural networks on new tasks in under-explored domains. Motivated by this goal, we study the problem of enabling users to discover the right neural operations given data from their specific domain. We introduce a search space of operations called XD-Operations that mimic the inductive bias of standard multi-channel convolutions while being much more expressive: we prove that it includes many named operations across multiple application areas. Starting with any standard backbone such as ResNet, we show how to transform it into a search space over XD-operations and how to traverse the space using a simple weight-sharing scheme. On a diverse set of taskssolving PDEs, distance prediction for protein folding, and music modeling-our approach consistently yields models with lower error than baseline networks and often even lower error than expert-designed domain-specific approaches. * denotes equal contribution. 35th Conference on Neural Information Processing Systems (NeurIPS 2021).

For agents

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

get_harvested_code_for_paper("2103.15798")
get_code_for_paper("2103.15798")
have("2103.15798")

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

curl https://syntology.ai/api/ran/2103.15798.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.15798.svg)](https://syntology.ai/paper/2103.15798)

Connect an agent — have() is free.