SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.04730 · 2020

Picket: Guarding Against Corrupted Data in Tabular Data during Learning and Inference

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; 0 honoured the contract we drafted, 0 violated it and 12 have no contract verdict; 1 failure 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/2006.04730.json · badge: /api/ran/2006.04730.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
rekords-uw/Picket community (classifier label)
classifier basis not recorded on the link
13 ran · 4 no recorded run
FunctionStatusWhere it lives
attention Ran
ran on a synthesized input
rekords-uw/Picket/picket/transformer/PicketNet.py
code served (permissive licence) · get_code("b1b1f3c94280b7cd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. 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 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
import numpy as np
import math

torch.manual_seed(0)

# Sparsemax class that the function uses
class Sparsemax(nn.Module):
    def forward(self, x):
        return F.softmax(x, dim=-1)

# Build tiny tensors
batch_size = 2
seq_len = 3
d_k = 4

query = torch.randn(batch_size, seq_len, d_k)
key = torch.randn(batch_size, seq_len, d_k)
value = torch.randn(batch_size, seq_len, d_k)

# Dropout layer
dropout_layer = nn.Dropout(p=0.1)

# Hard mask: shape matching scores (batch_size, seq_len, seq_len)
hard_mask = np.ones((batch_size, seq_len, seq_len))
hard_mask[0, 0, 1] = 0  # Mask out one position

ARGS = {
    "query": query,
    "key": key,
    "value": value,
    "mask": False,
    "dropout": dropout_layer,
    "hard_mask": hard_mask,
    "sparse": 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 — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
clones Ran
ran; our contract draft was wrong, not the code
rekords-uw/Picket/picket/transformer/PicketNet.py
pointer only (licence: MIT); the stored copy get_code serves, from DSE-MSU/R-transformer, is pointer only (licence: NONE) · get_code("b1bec2c4b1c9b491")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from DSE-MSU/R-transformer, 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.

dataPack Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTest.py
code served (permissive licence) · get_code("2bc0524c41027104")
How it ran

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

import numpy as np

# Create small numpy arrays with distinct dimensions for clarity
# X arrays: (batch_size, feature_dim)
# idx arrays: (batch_size,)

X_clean_correct = np.random.randn(2, 4)  # 2 samples, 4 features
idx_clean_correct = np.array([0, 1])

X_dirty_correct = np.random.randn(3, 4)  # 3 samples, 4 features
idx_dirty_correct = np.array([2, 3, 4])

X_dirty_wrong = np.random.randn(2, 4)  # 2 samples, 4 features
idx_dirty_wrong = np.array([5, 6])

ds = {
    'X_dirty_correct': X_dirty_correct,
    'idx_dirty_correct': idx_dirty_correct,
    'X_dirty_wrong': X_dirty_wrong,
    'idx_dirty_wrong': idx_dirty_wrong,
    'X_clean_correct': X_clean_correct,
    'idx_clean_correct': idx_clean_correct,
}

ARGS = {
    'ds': ds,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dataPackAdv Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTest.py
code served (permissive licence) · get_code("72947f687afde26a")
How it ran

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

import numpy as np

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

# Create ds_adv dictionary
# X_attack shape: (adv_size, attr_num, data_dim)
# The function checks that X_adv[:, :, 1:] is near zero
adv_size = 2
attr_num = 3
data_dim = 4

X_adv = np.zeros((adv_size, attr_num, data_dim))
X_adv[:, :, 0] = np.random.randn(adv_size, attr_num)  # Only first column has non-zero values

ds_adv = {
    'X_attack': X_adv
}

# Create ds_random dictionary
# X_clean_correct shape: (num_clean, attr_num, data_dim)
num_clean = 5
X_clean_correct = np.random.randn(num_clean, attr_num, data_dim)
idx_clean_correct = np.arange(num_clean)

ds_random = {
    'X_clean_correct': X_clean_correct,
    'idx_clean_correct': idx_clean_correct
}

ARGS = {
    'ds_adv': ds_adv,
    'ds_random': ds_random
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
flattern Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTrain.py
code served (permissive licence) · get_code("92be350945f9b822")
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, 12) 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 3D tensor with shape (batch=2, height=4, width=3)
# This allows us to test both first_dim=True and first_dim=False cases
X = torch.randn(2, 4, 3)

ARGS = {
    "X": X,
    "first_dim": 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, 12) — shape (2, 12) 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
getCleanDirtyMix Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTrain.py
code served (permissive licence) · get_code("08d29bc201ab1618")
How it ran

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

import numpy as np

# Create small numpy arrays for the function
# X: clean data array (batch_size=2, features=4)
X = np.array([
    [1.0, 2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0, 8.0]
], dtype=np.float32)

# X_dirty: dirty/corrupted data array (same shape as X)
X_dirty = np.array([
    [10.0, 20.0, 30.0, 40.0],
    [50.0, 60.0, 70.0, 80.0]
], dtype=np.float32)

# idx: clean indices array (batch_size=2)
idx = np.array([0, 1], dtype=np.int32)

# idx_dirty: dirty indices array (same shape as idx)
idx_dirty = np.array([10, 11], dtype=np.int32)

# eps_row: mixing probability parameter (scalar float between 0 and 1)
eps_row = 0.5

ARGS = {
    "X": X,
    "X_dirty": X_dirty,
    "idx": idx,
    "idx_dirty": idx_dirty,
    "eps_row": eps_row
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
getCleanPoisonMix Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTrain.py
code served (permissive licence) · get_code("0fb67b9d95e7e366")
How it ran

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

import numpy as np

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

# Create small test data
# X: clean data, shape (5, 4) - 5 samples, 4 features
X = np.random.randn(5, 4).astype(np.float32)

# X_dirty: poisoned data, shape (8, 4) - more samples than X for testing
X_dirty = np.random.randn(8, 4).astype(np.float32)

# y: clean labels, shape (5,)
y = np.array([0, 1, 0, 1, 0]).astype(np.int32)

# y_dirty: poisoned labels, shape (8,)
y_dirty = np.array([1, 1, 1, 0, 0, 1, 1, 0]).astype(np.int32)

# idx: indices for clean data, shape (5,)
idx = np.array([0, 1, 2, 3, 4]).astype(np.int32)

# eps_row: ratio parameter (must be between 0 and 1, exclusive)
# Using 0.2 means we'll take 20% / 80% = 0.25 * N of dirty samples
eps_row = 0.2

ARGS = {
    "X": X,
    "X_dirty": X_dirty,
    "y": y,
    "y_dirty": y_dirty,
    "idx": idx,
    "eps_row": eps_row
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
logit_fn Ran
ran on a synthesized input
rekords-uw/Picket/picket/rvae/model_utils.py
code served (permissive licence) · get_code("faa345f3ec52e705")
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 numpy as np

torch.manual_seed(0)

# The function logit_fn takes a single parameter x
# It expects x to be a tensor with values in (0, 1) range
# since it computes log(x) and log(1-x)

x = torch.tensor([[0.1, 0.5, 0.9, 0.3],
                   [0.2, 0.7, 0.4, 0.8]], dtype=torch.float32)

ARGS = {
    "x": x
}
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 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
nll_categ_global Ran
ran on a synthesized input
rekords-uw/Picket/picket/rvae/model_utils.py
code served (permissive licence) · get_code("ecf06b84d66c4983")
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, 1) 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.nn import functional as F

torch.manual_seed(0)

# Create tensor arguments
batch_size = 2
num_categories = 4

# categ_logp_feat: log probabilities from a categorical model
# Shape: (batch_size, num_categories)
categ_logp_feat = torch.randn(batch_size, num_categories)

# input_idx_feat: ground truth category indices
# Shape: (batch_size,)
input_idx_feat = torch.randint(0, num_categories, (batch_size,))

# cat_feat_size: number of categories
cat_feat_size = num_categories

# w: weights for robust loss (used when isRobust=True)
# Shape: (batch_size,)
w = torch.rand(batch_size)

ARGS = {
    'categ_logp_feat': categ_logp_feat,
    'input_idx_feat': input_idx_feat,
    'cat_feat_size': cat_feat_size,
    'isRobust': False,
    'w': w,
    'isClean': 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, 1) — shape (2, 1) 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
nll_gauss_global Ran
ran on a synthesized input
rekords-uw/Picket/picket/rvae/model_utils.py
code served (permissive licence) · get_code("4555a0d1dd7fd087")
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 small tensors with distinct dimensions for clarity
batch_size = 2
feat_dim = 4

# gauss_params: mean parameters, shape [batch_size * feat_dim]
gauss_params = torch.randn(batch_size * feat_dim)

# input_val_feat: input features, shape [batch_size * feat_dim]
input_val_feat = torch.randn(batch_size * feat_dim)

# logvar: log variance, shape [feat_dim]
logvar = torch.randn(feat_dim)

# w: weights for robust component, shape [batch_size * feat_dim]
w = torch.rand(batch_size * feat_dim)

# shape_feats: feature shape
shape_feats = [feat_dim]

# Other parameters
isRobust = True
isClean = False
std_0_scale = 2

ARGS = {
    "gauss_params": gauss_params,
    "input_val_feat": input_val_feat,
    "logvar": logvar,
    "isRobust": isRobust,
    "w": w,
    "std_0_scale": std_0_scale,
    "isClean": isClean,
    "shape_feats": shape_feats,
}
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
randomSelect Ran
ran on a synthesized input
rekords-uw/Picket/picket/filter/filtersTest.py
code served (permissive licence) · get_code("119b757f6c0660c4")
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 (5, 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

# Create a small numpy array for testing
X = np.random.randn(10, 4)

# Define size parameter
size = 5

ARGS = {
    "X": X,
    "size": size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (5, 4) — shape (5, 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
random_flip Ran
ran on a synthesized input
rekords-uw/Picket/picket/prepare/dataPrepare.py
code served (permissive licence) · get_code("b0f137dc86a5d3f1")
How it ran

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

import numpy as np

np.random.seed(0)

# Construct arguments for random_flip function
x = np.array([[1.0, 2.0, 3.0, 4.0],
              [5.0, 6.0, 7.0, 8.0]], dtype=np.float32)

idx = np.array([0, 1], dtype=np.int32)

eps = 0.5

# vec is a list of arrays for categorical features
vec = [
    np.array([[10.0, 20.0, 30.0, 40.0],
              [11.0, 21.0, 31.0, 41.0],
              [12.0, 22.0, 32.0, 42.0]], dtype=np.float32),
    np.array([[100.0, 200.0, 300.0, 400.0],
              [101.0, 201.0, 301.0, 401.0]], dtype=np.float32)
]

dtypes = ['categorical', 'numeric']

stds = [1.0, 2.0]

noise_std = 0.1

ARGS = {
    'x': x,
    'idx': idx,
    'eps': eps,
    'vec': vec,
    'dtypes': dtypes,
    'stds': stds,
    'noise_std': noise_std
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
standardize Ran
ran on a synthesized input
rekords-uw/Picket/picket/prepare/dataPrepare.py
code served (permissive licence) · get_code("f7ae8b3165f58eab")
How it ran

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

import numpy as np

# Construct arguments for standardize function
# X: shape (batch, features, 1) - contains numeric data
# vecs: list of arrays, one per feature, each with shape (n_vectors, 1)
# dtypes: list of strings indicating data type for each feature
# mean: optional, computed from X if None
# std: optional, computed from X if None

batch_size = 2
n_features = 4
n_vectors_per_feature = 3

# X: (batch_size, n_features, 1)
X = np.random.randn(batch_size, n_features, 1).astype(np.float32)

# vecs: list of n_features arrays, each (n_vectors_per_feature, 1)
vecs = [np.random.randn(n_vectors_per_feature, 1).astype(np.float32) for _ in range(n_features)]

# dtypes: list of strings indicating 'numeric' or other types
dtypes = ['numeric', 'numeric', 'numeric', 'numeric']

# mean and std: None (will be computed), or pre-computed
mean = None
std = None

ARGS = {
    'X': X,
    'vecs': vecs,
    'dtypes': dtypes,
    'mean': mean,
    'std': std
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
evaluation_phase Not yet run
unverified (harvested, no recorded run)
rekords-uw/Picket/picket/rvae/train_eval_models.py
code served (permissive licence) · get_code("f31567e2531cace6")
How it ran

no invocation recorded.

make_model Not yet run
unverified (harvested, no recorded run)
rekords-uw/Picket/picket/transformer/PicketNet.py
code served (permissive licence) · get_code("5cd95cc706535ebb")
How it ran

no invocation recorded.

repair_phase Not yet run
unverified (harvested, no recorded run)
rekords-uw/Picket/picket/rvae/train_eval_models.py
code served (permissive licence) · get_code("6242edd693182969")
How it ran

no invocation recorded.

training_phase Not yet run
unverified (harvested, no recorded run)
rekords-uw/Picket/picket/rvae/train_eval_models.py
code served (permissive licence) · get_code("6766b30d90dfd263")
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, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 1 the same month, 1 later) and 13 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

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

FunctionAlso present inEarliest-datedOthers named
clones 79 other papers
20 dated earlier, 3 the same month, 56 later · 11 only as an assembly brick
1411.4555 (2014-11)
Show and Tell: A Neural Image Caption Generator
1612.00563 (2016-12)
Self-critical Sequence Training for Image Captioning
1706.03762 (2017-06)
Attention Is All You Need
and 76 more not named here
logit_fn 1 other paper
1 dated earlier
1907.06671 (2019-07)
Robust Variational Autoencoders for Outlier Detection and Repair of Mixed-Type Data
nll_categ_global 1 other paper
1 dated earlier
1907.06671 (2019-07)
Robust Variational Autoencoders for Outlier Detection and Repair of Mixed-Type Data
nll_gauss_global 1 other paper
1 dated earlier
1907.06671 (2019-07)
Robust Variational Autoencoders for Outlier Detection and Repair of Mixed-Type Data

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

Data corruption is an impediment to modern machine learning deployments. Corrupted data can severely bias the learned model and can also lead to invalid inferences. We present, Picket, a simple framework to safeguard against data corruptions during both training and deployment of machine learning models over tabular data. For the training stage, Picket identifies and removes corrupted data points from the training data to avoid obtaining a biased model. For the deployment stage, Picket flags, in an online manner, corrupted query points to a trained machine learning model that due to noise will result in incorrect predictions. To detect corrupted data, Picket uses a self-supervised deep learning model for mixed-type tabular data, which we call PicketNet. To minimize the burden of deployment, learning a PicketNet model does not require any human-labeled data. Picket is designed as a plugin that can increase the robustness of any machine learning pipeline. We evaluate Picket on a diverse array of real-world data considering different corruption models that include systematic and adversarial noise during both training and testing. We show that Picket consistently safeguards against corrupted data during both training and deployment of various models ranging from SVMs to neural networks, beating a diverse array of competing methods that span from data quality validation models to robust outlier-detection models.

For agents

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

get_harvested_code_for_paper("2006.04730")
get_code_for_paper("2006.04730")
have("2006.04730")

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

curl https://syntology.ai/api/ran/2006.04730.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.04730.svg)](https://syntology.ai/paper/2006.04730)

Connect an agent — have() is free.