SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.02610 · ICCV · 2021

On the Robustness of Vision Transformers to Adversarial Examples

Kaleel Mahmood, Rigel Mahmood

arXiv · PDF · Open in the Atlas

Code that ran

11 of 15 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, 1 violated it and 0 have no contract verdict; 10 failures of our instrument; 4 no recorded run); 2 of them were also run on the host, not in the jail, by the peer-agreement lane that set their level, 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/2104.02610.json · badge: /api/ran/2104.02610.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
MetaMain/ViTRobust canonical (classifier label)
classifier basis: census
11 ran · 4 no recorded run
FunctionStatusWhere it lives
np2th Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/TransformerModels.py
code served (permissive licence) · get_code("7d8eaf5537f699e4")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 4, dtype family float, shape (5, 4, 2, 3) as a later run recorded it. The image it ran in was not recorded. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-05T00:36:31Z). 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)

# Create a small numpy array with shape HWIO (Height, Width, Input channels, Output channels)
# Using distinct dimensions: H=2, W=3, I=4, O=5
weights = np.random.randn(2, 3, 4, 5).astype(np.float32)

ARGS = {
    "weights": weights,
    "conv": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 (5, 4, 2, 3) — shape (5, 4, 2, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
swish Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/TransformerModels.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) 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 comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# So x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
GradientNormalizedByL1 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/AttackWrappersWhiteBoxP.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from queyrusi/pelta, is pointer only (licence: NONE) · get_code("71ded737bf803beb")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from queyrusi/pelta, 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.

conv1x1 Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/TransformerResNet.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from maloadba/mgenseg_2d, is pointer only (licence: NONE) · get_code("b0f916df39e78ae5")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from maloadba/mgenseg_2d, 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.

conv1x1 Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/BigTransferModels.py
code served (permissive licence) · get_code("ee80cf467065076b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned StdConv2d. 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)

ARGS = {
    'cin': 3,
    'cout': 16,
    'stride': 1,
    'bias': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type StdConv2d — a StdConv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a StdConv2d 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
conv3x3 Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/TransformerResNet.py
pointer only (licence: BSD-3-Clause); the stored copy get_code serves, from maloadba/mgenseg_2d, is pointer only (licence: NONE) · get_code("4e5bf13dbdc4f008")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-3-Clause), but get_code serves the stored copy, first harvested from maloadba/mgenseg_2d, 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/BigTransferModels.py
code served (permissive licence) · get_code("f01b8d3901f0b289")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned StdConv2d. 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)

ARGS = {
    'cin': 3,
    'cout': 16,
    'stride': 1,
    'groups': 1,
    'bias': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type StdConv2d — a StdConv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a StdConv2d 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
tf2th Ran
ran, violated the contract we drafted
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/BigTransferModels.py
code served (permissive licence) · get_code("7023eb86801bb3a3")
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 (6, 5, 3, 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)

# Create a 4D numpy array in HWIO format (Height, Width, Input channels, Output channels)
# Using distinct dimensions: H=3, W=4, I=5, O=6
conv_weights = np.random.randn(3, 4, 5, 6).astype(np.float32)

ARGS = {
    "conv_weights": conv_weights
}
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 (6, 5, 3, 4) — shape (6, 5, 3, 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
validateD Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/DataManagerPytorch.py
code served (permissive licence) · get_code("72cd943b81f2517e")
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. 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
import torch.nn as nn

torch.manual_seed(0)

# Create a simple model that accepts input and produces class predictions
class SimpleClassifier(nn.Module):
    def __init__(self, input_dim=4, num_classes=3):
        super(SimpleClassifier, self).__init__()
        self.fc = nn.Linear(input_dim, num_classes)
    
    def forward(self, x):
        return self.fc(x)

# Create a simple dataset as a list of (input, target) tuples
class SimpleDataset:
    def __init__(self, num_samples=4, input_dim=4, num_classes=3):
        self.data = []
        for _ in range(num_samples):
            input_tensor = torch.randn(1, input_dim)
            target = torch.tensor(torch.randint(0, num_classes, (1,)).item())
            self.data.append((input_tensor, target))
    
    def __iter__(self):
        return iter(self.data)
    
    def __len__(self):
        return len(self.data)

# Create a data loader wrapper
class SimpleDataLoader:
    def __init__(self, dataset):
        self.dataset = dataset
    
    def __iter__(self):
        # Batch the data: return batches of size 2
        data_list = list(self.dataset)
        for i in range(0, len(data_list), 2):
            batch = data_list[i:i+2]
            # Stack inputs and targets
            inputs = torch.cat([item[0] for item in batch], dim=0)
            targets = torch.stack([item[1] for item in batch])
            yield (inputs, targets)
    
    def __len__(self):
        return len(self.dataset)

# Instantiate components
model = SimpleClassifier(input_dim=4, num_classes=3)
dataset = SimpleDataset(num_samples=4, input_dim=4, num_classes=3)
valLoader = SimpleDataLoader(dataset)

ARGS = {
    'valLoader': valLoader,
    'model': model,
    'device': None
}
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
validateDA Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/DataManagerPytorch.py
code served (permissive licence) · get_code("7d6d811b990ecc66")
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 1, dtype family float, shape (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
import torch.nn as nn

torch.manual_seed(0)

# Build a minimal model that outputs class logits
class SimpleModel(nn.Module):
    def __init__(self, num_classes=4):
        super().__init__()
        self.linear = nn.Linear(5, num_classes)
    
    def forward(self, x):
        return self.linear(x)

# Create a minimal data loader (list of (input, target) tuples)
# Each input is shape (batch_size, 5), each target is a tensor of class indices
valLoader = [
    (torch.randn(2, 5), torch.tensor([0, 1])),
    (torch.randn(2, 5), torch.tensor([2, 3])),
]

# Add dataset attribute so len(valLoader.dataset) works
class SimpleDataset:
    def __init__(self, size):
        self.size = size
    
    def __len__(self):
        return self.size

# Wrap loader with dataset
class WrappedLoader:
    def __init__(self, batches, dataset_size):
        self.batches = batches
        self.dataset = SimpleDataset(dataset_size)
    
    def __iter__(self):
        return iter(self.batches)
    
    def __len__(self):
        return len(self.batches)

valLoader = WrappedLoader(valLoader, 4)

model = SimpleModel(num_classes=4)

device = None

ARGS = {
    "valLoader": valLoader,
    "model": model,
    "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 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
validateT Ran
ran; our contract draft was wrong, not the code
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/DataManagerPytorch.py
code served (permissive licence) · get_code("4d607f4733726c9d")
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
import math

torch.manual_seed(0)

# Create a simple model that takes input and produces class logits
class SimpleClassifier(nn.Module):
    def __init__(self, input_dim, num_classes):
        super().__init__()
        self.fc = nn.Linear(input_dim, num_classes)
    
    def forward(self, x):
        return self.fc(x)

# Create small tensors for testing
num_samples = 2
input_dim = 4
num_classes = 3

xData = torch.randn(num_samples, input_dim)
yData = torch.tensor([0, 2])

model = SimpleClassifier(input_dim, num_classes)

ARGS = {
    "xData": xData,
    "yData": yData,
    "model": model,
    "batchSize": None
}
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
GetAttention Not yet run
unverified (harvested, no recorded run)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/AttackWrappersSAGA.py
code served (permissive licence) · get_code("efa84ad482883af5")
How it ran

no invocation recorded.

resnet1001 Not yet run
unverified (harvested, no recorded run)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/ResNetPytorch.py
code served (permissive licence) · get_code("975c33663559b367")
How it ran

no invocation recorded.

resnet164 Not yet run
unverified (harvested, no recorded run)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/ResNetPytorch.py
code served (permissive licence) · get_code("55fcf6ec82c11e08")
How it ran

no invocation recorded.

resnet56 Not yet run
unverified (harvested, no recorded run)
MetaMain/ViTRobust/VisionTransformersRobustness/VisionTransformersRobustness/ResNetPytorch.py
code served (permissive licence) · get_code("881ff295da65733e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 9 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 2 the same month, 9 later) and 4 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
swish 206 other papers
63 dated earlier, 3 the same month, 137 later, 3 same year, order not determinable · 44 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1805.12114 (2018-05)
Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
1809.01357 (2018-09)
Zero Shot Learning for Code Education: Rubric Sampling with Deep Learning Inference
and 203 more not named here
np2th 35 other papers
5 dated earlier, 1 the same month, 29 later · 5 only as an assembly brick
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2102.04306 (2021-02)
TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation
2103.07976 (2021-03)
TransFG: A Transformer Architecture for Fine-grained Recognition
and 32 more not named here
conv1x1 33 other papers
4 dated earlier, 29 later · 3 only as an assembly brick
1912.11370 (2019-12)
Big Transfer (BiT): General Visual Representation Learning
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2102.04306 (2021-02)
TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation
and 30 more not named here
conv3x3 33 other papers
4 dated earlier, 29 later · 3 only as an assembly brick
1912.11370 (2019-12)
Big Transfer (BiT): General Visual Representation Learning
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2102.04306 (2021-02)
TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation
and 30 more not named here
tf2th 17 other papers
1 dated earlier, 16 later
1912.11370 (2019-12)
Big Transfer (BiT): General Visual Representation Learning
2106.01127 (2021-06)
Towards Robust Classification Model by Counterfactual and Invariant Data Generation
2108.02833 (2021-08)
Elaborative Rehearsal for Zero-shot Action Recognition
and 14 more not named here
GradientNormalizedByL1 1 other paper
1 later
2309.07197 (2023-09)
Mitigating Adversarial Attacks in Federated Learning with Trusted Execution Environments
resnet1001 1 other paper
1 later
2309.07197 (2023-09)
Mitigating Adversarial Attacks in Federated Learning with Trusted Execution Environments
resnet164 1 other paper
1 later
2309.07197 (2023-09)
Mitigating Adversarial Attacks in Federated Learning with Trusted Execution Environments
resnet56 1 other paper
1 later
2309.07197 (2023-09)
Mitigating Adversarial Attacks in Federated Learning with Trusted Execution Environments

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

Recent advances in attention-based networks have shown that Vision Transformers can achieve state-of-the-art or near state-of-the-art results on many image classification tasks. This puts transformers in the unique position of being a promising alternative to traditional convolutional neural networks (CNNs). While CNNs have been carefully studied with respect to adversarial attacks, the same cannot be said of Vision Transformers. In this paper, we study the robustness of Vision Transformers to adversarial examples. Our analyses of transformer security is divided into three parts. First, we test the transformer under standard whitebox and black-box attacks. Second, we study the transferability of adversarial examples between CNNs and transformers. We show that adversarial examples do not readily transfer between CNNs and transformers. Based on this finding, we analyze the security of a simple ensemble defense of CNNs and transformers. By creating a new attack, the self-attention blended gradient attack, we show that such an ensemble is not secure under a white-box adversary. However, under a black-box adversary, we show that an ensemble can achieve unprecedented robustness without sacrificing clean accuracy. Our analysis for this work is done using six types of white-box attacks and two types of black-box attacks. Our study encompasses multiple Vision Transformers, Big Transfer Models and CNN architectures trained on CIFAR-10, CIFAR-100 and ImageNet.

For agents

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

get_harvested_code_for_paper("2104.02610")
get_code_for_paper("2104.02610")
have("2104.02610")

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

curl https://syntology.ai/api/ran/2104.02610.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.02610.svg)](https://syntology.ai/paper/2104.02610)

Connect an agent — have() is free.