SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2301.03580 · 2023

Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling

arXiv · PDF · Open in the Atlas

Code that ran

13 of 14 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; 2 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 3 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2301.03580.json · badge: /api/ran/2301.03580.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
keyu-tian/spark canonical (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
Factral/UnMix-NeRF community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
FunctionStatusWhere it lives
component_x Ran
ran on a synthesized input
Factral/UnMix-NeRF/unmixnerf/utils/spec_to_rgb.py
code served (permissive licence) · get_code("a883944500f3f33d")
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 1, dtype family float, shape (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

def component_x(x):
    def g(x, alpha, mu, sigma1, sigma2):
        sigma = np.clip((x < mu) * sigma1 + (x >= mu) * sigma2, a_min=1e-06, a_max=None)
        return alpha * np.exp((x - mu) ** 2 / (-2 * sigma ** 2))
    
    return g(x, 1.056, 5998, 379, 310) + g(x, 0.362, 4420, 160, 267) + g(x, -0.065, 5011, 204, 262)

# Create test input: small numpy array
x = np.array([4000, 5000, 6000, 7000], dtype=np.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 — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (4,) — shape (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
component_y Ran
ran on a synthesized input
Factral/UnMix-NeRF/unmixnerf/utils/spec_to_rgb.py
code served (permissive licence) · get_code("8da57cf56b2fdcf1")
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 1, dtype family float, shape (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

def component_y(x):
    def g(x, alpha, mu, sigma1, sigma2):
        sigma = np.clip((x < mu) * sigma1 + (x >= mu) * sigma2, a_min=1e-06, a_max=None)
        return alpha * np.exp((x - mu) ** 2 / (-2 * sigma ** 2))
    
    return g(x, 0.821, 5688, 469, 405) + g(x, 0.286, 5309, 163, 311)

# Create test data for x parameter
x = np.array([5000, 5500, 6000, 6500], dtype=np.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 — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (4,) — shape (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
g Ran
ran on a synthesized input
Factral/UnMix-NeRF/unmixnerf/utils/spec_to_rgb.py
code served (permissive licence) · get_code("558f6d7e47654006")
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 1, dtype family float, shape (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

# Analyze the function g:
# - x: used in arithmetic operations, needs to be numeric array-like
# - alpha: scalar, multiplied with exponential result
# - mu: scalar, used in comparison and arithmetic
# - sigma1: scalar, used in conditional selection
# - sigma2: scalar, used in conditional selection

# Create small test arrays
x = np.array([0.5, 1.5, 2.5, 3.5])
alpha = 0.8
mu = 2.0
sigma1 = 0.5
sigma2 = 1.0

ARGS = {
    "x": x,
    "alpha": alpha,
    "mu": mu,
    "sigma1": sigma1,
    "sigma2": sigma2
}
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 1, dtype family float, shape (4,) — shape (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
get_num_layer_layer_wise Ran
ran, honoured the contract we drafted
keyu-tian/spark/downstream_mmdet/mmcv_custom/layer_decay_optimizer_constructor.py
code served (permissive licence) · get_code("aa8d49041d9f8442")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function get_num_layer_layer_wise takes two parameters:
# 1. var_name: a string representing a variable name in a backbone model
# 2. num_max_layer: an integer with default value 12

# Let's create valid test cases for var_name that exercise different branches

ARGS = {
    'var_name': 'backbone.cls_token',
    'num_max_layer': 12
}
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 int, ndim 0, dtype family int, 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_num_layer_stage_wise Ran
ran, honoured the contract we drafted
keyu-tian/spark/downstream_mmdet/mmcv_custom/layer_decay_optimizer_constructor.py
code served (permissive licence) · get_code("be134d6ef63b5192")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function takes two parameters: var_name (string) and num_max_layer (int)
# It doesn't actually use the objects themselves, just string matching and integer arithmetic
# So we can pass simple test values

ARGS = {
    'var_name': 'backbone.stages.1.conv',
    'num_max_layer': 4
}
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 int, ndim 0, dtype family int, 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_param_groups Ran
ran on a synthesized input
keyu-tian/spark/downstream_imagenet/lr_decay.py
code served (permissive licence) · get_code("ad373c0953d230ed")
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
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.layer1_weight = nn.Parameter(torch.randn(4, 3))
        self.layer1_bias = nn.Parameter(torch.randn(4))
        self.layer2_weight = nn.Parameter(torch.randn(5, 4))
        self.layer2_bias = nn.Parameter(torch.randn(5))
        self.norm_scale = nn.Parameter(torch.randn(5))
    
    def forward(self, x):
        return x

model = SimpleModel()

ARGS = {
    'model': model,
    'nowd_keys': ('norm',),
    'lr_scale': 0.5
}
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
get_weights_spectral Ran
ran on a synthesized input
Factral/UnMix-NeRF/unmixnerf/unmix_renderer.py
code served (permissive licence) · get_code("36ae228f14629970")
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 3, dtype family float, shape (2, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for deltas and densities
# deltas: shape (batch=2, num_samples=4, channels=3)
# densities: shape (batch=2, num_samples=4, channels=3)
deltas = torch.rand(2, 4, 3)
densities = torch.rand(2, 4, 3)

ARGS = {
    "deltas": deltas,
    "densities": densities,
}
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 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_checkpoint Ran
ran on a synthesized input
keyu-tian/spark/downstream_imagenet/util.py
code served (permissive licence) · get_code("7475c50600ea1392")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import torch
import torch.nn as nn
import tempfile

torch.manual_seed(0)

# Create a minimal model
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create a minimal EMA module
class MinimalEMA(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Create a minimal optimizer
model = MinimalModel()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)

# Create a checkpoint file
checkpoint_data = {
    'epoch': 5,
    'performance_desc': 'test_performance',
    'module': model.state_dict(),
    'optimizer': optimizer.state_dict(),
    'ema': MinimalEMA().state_dict(),
    'is_pretrain': False,
}

# Write checkpoint to a temporary file
temp_dir = tempfile.gettempdir()
checkpoint_path = os.path.join(temp_dir, 'test_checkpoint.pth')
torch.save(checkpoint_data, checkpoint_path)

# Create model and EMA instances for loading
model_without_ddp = MinimalModel()
ema_module = MinimalEMA()

# Create optimizer for loading
optimizer_for_load = torch.optim.SGD(model_without_ddp.parameters(), lr=0.01)

ARGS = {
    'resume_from': checkpoint_path,
    'model_without_ddp': model_without_ddp,
    'ema_module': ema_module,
    'optimizer': optimizer_for_load,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
lr_factor_func Ran
ran on a synthesized input
keyu-tian/spark/downstream_d2/lr_decay.py
code served (permissive licence) · get_code("800af14f9c740324")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

def lr_factor_func(para_name: str, is_resnet50, dec: float, debug=False) -> float:
    if dec == 0:
        dec = 1.0
    N = 5 if is_resnet50 else 11
    if '.stem.' in para_name:
        layer_id = 0
    elif '.res' in para_name:
        ls = para_name.split('.res')[1].split('.')
        if ls[0].isnumeric() and ls[1].isnumeric():
            stage_id, block_id = (int(ls[0]), int(ls[1]))
            if stage_id == 2:
                layer_id = 1
            elif stage_id == 3:
                layer_id = 2
            elif stage_id == 4:
                layer_id = 3 + block_id // 3
            else:
                layer_id = N
        else:
            assert para_name.startswith('roi_heads.res5.norm.')
            layer_id = N + 1
    else:
        layer_id = N + 1
    exp = N + 1 - layer_id
    return f'{dec:g} ** {exp}' if debug else dec ** exp

torch.manual_seed(0)

ARGS = {
    'para_name': '.stem.conv1.weight',
    'is_resnet50': True,
    'dec': 0.5,
    'debug': 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 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
lr_wd_annealing Ran
ran on a synthesized input
keyu-tian/spark/downstream_imagenet/lr_decay.py
code served (permissive licence) · get_code("f921a5666329b4eb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import torch.optim as optim
import math

torch.manual_seed(0)

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

# Create an optimizer
optimizer = optim.SGD(
    [
        {'params': model.weight, 'lr_scale': 1.0, 'weight_decay_scale': 1.0},
        {'params': model.bias, 'lr_scale': 0.5, 'weight_decay_scale': 2.0}
    ],
    lr=0.1,
    weight_decay=0.01
)

ARGS = {
    'optimizer': optimizer,
    'peak_lr': 0.1,
    'wd': 0.01,
    'cur_it': 50,
    'wp_it': 100.5,
    'max_it': 1000
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mixup_target Ran
ran; our contract draft was wrong, not the code
keyu-tian/spark/downstream_imagenet/mixup.py
pointer only (licence: MIT); the stored copy get_code serves, from terrypei/efficientvmamba, is pointer only (licence: NONE) · get_code("8733e0d87f232d11")
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 terrypei/efficientvmamba, 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.

one_hot Ran
ran; our contract draft was wrong, not the code
keyu-tian/spark/downstream_imagenet/mixup.py
code served (permissive licence) · get_code("4728dc2025227bba")
How it ran

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

import torch

torch.manual_seed(0)

# Create a small input tensor with class indices
# Shape: (2,) - batch of 2 samples with class indices
x = torch.tensor([0, 2], dtype=torch.long)

# Number of classes
num_classes = 4

# On value for one-hot encoding
on_value = 1.0

# Off value for one-hot encoding
off_value = 0.0

# Device - use CPU only
device = 'cpu'

ARGS = {
    'x': x,
    'num_classes': num_classes,
    'on_value': on_value,
    'off_value': off_value,
    'device': device
}
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
rand_bbox Ran
ran; our contract draft was wrong, not the code
keyu-tian/spark/downstream_imagenet/mixup.py
pointer only (licence: MIT); the stored copy get_code serves, from terrypei/efficientvmamba, is pointer only (licence: NONE) · get_code("3913c620f6926f42")
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 terrypei/efficientvmamba, 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.

create_classification_dataset Not yet run
unverified (harvested, no recorded run)
keyu-tian/spark/downstream_imagenet/data.py
code served (permissive licence) · get_code("0e865e01e71cf326")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 5 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 5 later) and 9 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
rand_bbox 14 other papers
4 dated earlier, 10 later
2109.03508 (2021-09)
RepNAS: Searching for Efficient Re-parameterizing Blocks
2111.06377 (2021-11)
Masked Autoencoders Are Scalable Vision Learners
2206.13559 (2022-06)
ST-Adapter: Parameter-Efficient Image-to-Video Transfer Learning
and 11 more not named here
one_hot 12 other papers
4 dated earlier, 8 later · 1 only as an assembly brick
1710.09412 (2017-10)
mixup: BEYOND EMPIRICAL RISK MINIMIZATION
2109.03508 (2021-09)
RepNAS: Searching for Efficient Re-parameterizing Blocks
2212.00776 (2022-12)
ResFormer: Scaling ViTs with Multi-Resolution Training
and 9 more not named here
mixup_target 7 other papers
2 dated earlier, 5 later
2109.03508 (2021-09)
RepNAS: Searching for Efficient Re-parameterizing Blocks
2212.00776 (2022-12)
ResFormer: Scaling ViTs with Multi-Resolution Training
2305.02507 (2023-05)
Stimulative Training++: Go Beyond The Performance Limits of Residual Networks
and 4 more not named here
get_num_layer_stage_wise 4 other papers
4 later
2302.06052 (2023-02)
CEDNet: A Cascade Encoder-Decoder Network for Dense Prediction
2312.10758 (2023-12)
SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation
2403.05369 (2024-03)
Frequency-Adaptive Dilated Convolution for Semantic Segmentation
and 1 more not named here
get_num_layer_layer_wise 3 other papers
3 later
2302.06052 (2023-02)
CEDNet: A Cascade Encoder-Decoder Network for Dense Prediction
2312.10758 (2023-12)
SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation
2403.17409 (2024-03)
Neural Clustering based Visual Representation Learning

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

We identify and overcome two key obstacles in extending the success of BERT-style pre-training, or the masked image modeling, to convolutional networks (convnets): (i) convolution operation cannot handle irregular, random-masked input images; (ii) the single-scale nature of BERT pre-training is inconsistent with convnet's hierarchical structure. For (i), we treat unmasked pixels as sparse voxels of 3D point clouds and use sparse convolution to encode. This is the first use of sparse convolution for 2D masked modeling. For (ii), we develop a hierarchical decoder to reconstruct images from multi-scale encoded features. Our method called Sparse masKed modeling (SparK) is general: it can be used directly on any convolutional model without backbone modifications. We validate it on both classical (ResNet) and modern (ConvNeXt) models: on three downstream tasks, it surpasses both state-of-the-art contrastive learning and transformer-based masked modeling by similarly large margins (around +1.0%). Improvements on object detection and instance segmentation are more substantial (up to +3.5%), verifying the strong transferability of features learned. We also find its favorable scaling behavior by observing more gains on larger models. All this evidence reveals a promising future of generative pre-training on convnets. Codes and models are released at https://github.com/keyu-tian/SparK.

For agents

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

get_harvested_code_for_paper("2301.03580")
get_code_for_paper("2301.03580")
have("2301.03580")

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

curl https://syntology.ai/api/ran/2301.03580.json

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

[![Syntology run record](https://syntology.ai/api/ran/2301.03580.svg)](https://syntology.ai/paper/2301.03580)

Connect an agent — have() is free.