SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.00345 · 2020

Editable Neural Networks

arXiv · PDF · Open in the Atlas

Code that ran

13 of 17 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 10 have no contract verdict; 2 failures of our instrument; 4 no recorded run), measured 2026-11-21; 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/2004.00345.json · badge: /api/ran/2004.00345.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
xtinkt/editable canonical (classifier label)
classifier basis: census
12 ran · 4 no recorded run
editable-ICLR2020/editable canonical (classifier label)
classifier basis: census
1 ran · 0 no recorded run
FunctionStatusWhere it lives
check_numpy Ran
ran, honoured the contract we drafted
xtinkt/editable/lib/utils/data.py
code served (permissive licence) · get_code("05c9ea23ab69917c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (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 numpy as np
import torch

torch.manual_seed(0)

# The function check_numpy takes a single parameter 'x'
# It can accept: torch.Tensor, list, numpy array, or anything convertible to numpy array
# Let's test with a torch.Tensor as that's the most complex case

x_tensor = torch.randn(2, 4)

ARGS = {
    "x": x_tensor
}
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 ndarray, 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
contrastive_cross_entropy Ran
ran on a synthesized input
xtinkt/editable/lib/loss.py
code served (permissive licence) · get_code("29be574face2d8d5")
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 expects:
# - logits: a tensor of shape (batch_size, num_classes) with raw logit scores
# - target: a tensor of shape (batch_size,) with class indices
# - margin: a float (optional, defaults to 0.0)

batch_size = 2
num_classes = 4

logits = torch.randn(batch_size, num_classes)
target = torch.tensor([0, 2])
margin = 0.5

ARGS = {
    "logits": logits,
    "target": target,
    "margin": margin,
}
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
copy_and_replace Ran
ran; our contract draft was wrong, not the code
xtinkt/editable/lib/utils/copy_and_replace.py
code served (permissive licence) · get_code("7edf4c0dc6d1125d")
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 NestedContainer. 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 copy import deepcopy

torch.manual_seed(0)

# Create some simple objects to use in the test
class SimpleObject:
    def __init__(self, value):
        self.value = value

class NestedContainer:
    def __init__(self):
        self.obj1 = SimpleObject(42)
        self.obj2 = SimpleObject(100)
        self.tensor = torch.randn(2, 4)
        self.list = [1, 2, 3, SimpleObject(999)]

# Set up test objects
original_obj1 = SimpleObject(10)
original_obj2 = SimpleObject(20)
original_nested = NestedContainer()

# Create replacement object
replacement_obj = SimpleObject(999)

# Create do_not_copy list with some objects
do_not_copy_items = [original_obj1, original_obj2]

# Create replace dictionary mapping old objects to new ones
replace_dict = {
    original_nested.obj1: replacement_obj,
    original_nested.list[-1]: SimpleObject(1111)
}

ARGS = {
    "original": original_nested,
    "replace": replace_dict,
    "do_not_copy": do_not_copy_items
}
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 NestedContainer — a NestedContainer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NestedContainer 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
dot Ran
ran on a synthesized input
xtinkt/editable/lib/utils/basic.py
code served (permissive licence) · get_code("c2234bf157dfacb3")
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 4, dtype family float, shape (2, 3, 5, 6) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function performs matrix multiplication on reshaped tensors
# x.view(-1, x.shape[-1]) reshapes x to (batch_size, last_dim)
# y.view(y.shape[0], -1) reshapes y to (first_dim, flattened_rest)
# Then multiplies them and reshapes back to (*x.shape[:-1], *y.shape[1:])

# Create two tensors with distinct dimensions for clarity
# x: shape (2, 3, 4) - batch=2, intermediate=3, last_dim=4
# y: shape (4, 5, 6) - first_dim=4, rest=(5, 6)

x = torch.randn(2, 3, 4)
y = torch.randn(4, 5, 6)

ARGS = {
    "x": x,
    "y": y,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 3, 5, 6) — shape (2, 3, 5, 6) 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_learning_rate Ran
ran on a synthesized input
xtinkt/editable/lib/utils/learning_rate.py
code served (permissive licence) · get_code("ea79a3921665cbad")
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 torch.nn as nn
from torch.optim import SGD

torch.manual_seed(0)

# Create a simple model
model = nn.Linear(4, 3)

# Create an optimizer with a learning rate
optimizer = SGD(model.parameters(), lr=0.01)

# Define arguments for get_learning_rate
ARGS = {
    'optimizer': optimizer
}
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 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
get_updated_model Ran
ran on a synthesized input
xtinkt/editable/lib/utils/ingraph_update.py
code served (permissive licence) · get_code("a7996959cff5e426")
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 SimpleModel. 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 as nn

torch.manual_seed(0)

# Define a minimal model
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
model = SimpleModel()

# Create input and target for computing loss
x = torch.randn(2, 4)
target = torch.randn(2, 3)

# Compute loss
output = model(x)
loss = nn.MSELoss()(output, target)

# Define parameters to update
parameters = list(model.parameters())

# Define gradients (can be None if loss is provided)
gradients = None

# Stub for copy_and_replace function that the main function calls
def copy_and_replace(model, updates, do_not_copy):
    """Minimal stub implementation for copy_and_replace"""
    import copy
    new_model = copy.deepcopy(model)
    return new_model

# Make copy_and_replace available in the module scope
import sys
current_module = sys.modules[__name__]
setattr(current_module, 'copy_and_replace', copy_and_replace)

ARGS = {
    'model': model,
    'loss': loss,
    'gradients': gradients,
    'parameters': parameters,
    'detach': False,
    'learning_rate': 0.01,
    'allow_unused': 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 SimpleModel — a SimpleModel output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleModel 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
gumbel_sigmoid Ran
ran on a synthesized input
xtinkt/editable/lib/utils/gumbel.py
code served (permissive licence) · get_code("9d12c26ca7e41a01")
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)

# Create a small logits tensor with shape (2, 4)
logits = torch.randn(2, 4, dtype=torch.float32)

# Define ARGS mapping all parameters of gumbel_sigmoid
ARGS = {
    "logits": logits,
    "tau": 1.0,
    "noise": 1.0,
    "hard": 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
gumbel_softmax Ran
ran on a synthesized input
xtinkt/editable/lib/utils/gumbel.py
code served (permissive licence) · get_code("57aebddd2139800f")
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)

def to_one_hot(indices, depth):
    """Helper function to convert indices to one-hot encoding"""
    one_hot = torch.zeros(indices.shape + (depth,), dtype=indices.dtype)
    one_hot.scatter_(-1, indices.unsqueeze(-1), 1)
    return one_hot

# Construct arguments for gumbel_softmax
logits = torch.randn(2, 4)  # batch_size=2, num_classes=4
dim = -1
tau = 1.5
noise = 1.0
hard = False

ARGS = {
    "logits": logits,
    "dim": dim,
    "tau": tau,
    "noise": noise,
    "hard": hard,
}
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
init_normalized_ Ran
ran on a synthesized input
xtinkt/editable/lib/utils/init.py
code served (permissive licence) · get_code("c6a6b624a3d2bfec")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small tensor to initialize
x = torch.randn(2, 4)

# Create an initialization function (using default normal_)
init_func = nn.init.normal_

ARGS = {
    "x": x,
    "init_": init_func,
    "dim": -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 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
kl_distill_loss Ran
ran on a synthesized input
xtinkt/editable/lib/loss.py
code served (permissive licence) · get_code("666418dcb08b4e03")
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)

# Create small tensors for the function parameters
# logits: raw model outputs, shape (batch_size, num_classes)
logits = torch.randn(2, 4)

# ref_probs: reference probabilities (e.g., from teacher model), shape (batch_size, num_classes)
# These should be valid probability distributions (sum to 1 across classes)
ref_probs = F.softmax(torch.randn(2, 4), dim=-1)

ARGS = {
    "logits": logits,
    "ref_probs": ref_probs,
}
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
straight_through_grad Ran
ran on a synthesized input
xtinkt/editable/lib/utils/basic.py
code served (permissive licence) · get_code("85ab18dba2e770d6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple function that takes a tensor and returns a modified tensor
def simple_function(x, scale=1.0):
    """A simple function that scales and adds to input"""
    return x * scale + 1.0

# Test with single tensor input/output
def test_function_single(x, offset=0.5):
    """Function with single tensor output"""
    return x + offset

# Test with multiple tensor inputs/outputs
def test_function_multiple(x, y, scale=2.0):
    """Function with multiple tensor inputs and outputs"""
    return x * scale, y * scale

# Create sample tensors
input_tensor = torch.randn(2, 4)
input_tensor_1 = torch.randn(2, 3)
input_tensor_2 = torch.randn(2, 5)

ARGS = {
    "function": simple_function,
    "kwargs": {"scale": 2.0}
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
threshold_mse Ran
ran on a synthesized input
xtinkt/editable/lib/loss.py
code served (permissive licence) · get_code("dca6e052a5239e27")
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)

# Create small tensors for predictions and targets
predictions = torch.randn(2, 4)  # batch size 2, dimension 4
targets = torch.randn(2, 4)      # batch size 2, dimension 4

# Define arguments
ARGS = {
    "predictions": predictions,
    "targets": targets,
    "threshold": 0.5,
    "reduction_axes": (1,)  # Sum along dimension 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 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
to_one_hot Ran
ran; our contract draft was wrong, not the code
editable-ICLR2020/editable/lib/utils/basic.py
pointer only (licence: NONE) · get_code("c169a4c49b79cc3e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, editable-ICLR2020/editable 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.

classification_error Not yet run
unverified (harvested, no recorded run)
xtinkt/editable/lib/evaluate.py
code served (permissive licence) · get_code("87bcbca994f11504")
How it ran

no invocation recorded.

download Not yet run
unverified (harvested, no recorded run)
xtinkt/editable/lib/utils/data.py
code served (permissive licence) · get_code("4df0769f954f1705")
How it ran

no invocation recorded.

eval_func Not yet run
unverified (harvested, no recorded run)
xtinkt/editable/lib/evaluate.py
code served (permissive licence) · get_code("306656fe5dd0107d")
How it ran

no invocation recorded.

process_in_chunks Not yet run
unverified (harvested, no recorded run)
xtinkt/editable/lib/utils/data.py
code served (permissive licence) · get_code("49f6bbe3d00bcf2a")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 4 later) and 11 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
download 3 other papers
1 dated earlier, 2 later
1909.06312 (2019-09)
Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data
2209.15409 (2022-09)
Higher-order Neural Additive Models: An Interpretable Machine Learning Model with Feature Interactions
2307.04755 (2023-07)
Information decomposition in complex systems via machine learning
to_one_hot 3 other papers
1 dated earlier, 2 later
1909.06312 (2019-09)
Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data
2207.08548 (2022-07)
GANDALF: Gated Adaptive Network for Deep Automated Learning of Features
2407.09790 (2024-07)
Team up GBDTs and DNNs: Advancing Efficient and Effective Tabular Prediction with Tree-hybrid MLPs
check_numpy 2 other papers
1 dated earlier, 1 later
1909.06312 (2019-09)
Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data
2106.01613 (2021-06)
NODE-GAM: Neural Generalized Additive Model for Interpretable Deep Learning
process_in_chunks 2 other papers
1 dated earlier, 1 later
1909.06312 (2019-09)
Neural Oblivious Decision Ensembles for Deep Learning on Tabular Data
2106.01613 (2021-06)
NODE-GAM: Neural Generalized Additive Model for Interpretable Deep Learning
copy_and_replace 1 other paper
1 dated earlier
1909.04630 (2019-09)
Meta-Learning with Implicit Gradients
get_learning_rate 1 other paper
1 dated earlier
1806.03198 (2018-06)
Spreading vectors for similarity search

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 2026-11-21.

Repositories linked to this paper

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

Abstract

These days deep neural networks are ubiquitously used in a wide range of tasks, from image classification and machine translation to face identification and self-driving cars. In many applications, a single model error can lead to devastating financial, reputational and even life-threatening consequences. Therefore, it is crucially important to correct model mistakes quickly as they appear. In this work, we investigate the problem of neural network editing $-$ how one can efficiently patch a mistake of the model on a particular sample, without influencing the model behavior on other samples. Namely, we propose Editable Training, a model-agnostic training technique that encourages fast editing of the trained model. We empirically demonstrate the effectiveness of this method on large-scale image classification and machine translation tasks.

For agents

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

get_harvested_code_for_paper("2004.00345")
get_code_for_paper("2004.00345")
have("2004.00345")

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

curl https://syntology.ai/api/ran/2004.00345.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.00345.svg)](https://syntology.ai/paper/2004.00345)

Connect an agent — have() is free.