SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1908.08016 · 2019

Testing Robustness Against Unforeseen Adversaries

arXiv · PDF · Open in the Atlas

Code that ran

16 of 18 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 13 have no contract verdict; 2 failures of our instrument; 2 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/1908.08016.json · badge: /api/ran/1908.08016.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
ddkang/advex-uar canonical (classifier label)
classifier basis: census
13 ran · 2 no recorded run
centerforaisafety/adversarial-corruptions canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
compute_uar Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_uar.py
code served (permissive licence) · get_code("397083ca14c129e6")
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 numpy as np

torch.manual_seed(0)

# Construct atas: list of attack configurations
# Each ata is a list where ata[0] is [attack_type, epsilon]
# and ata[1] is a count/weight value
atas = [
    [["fgsm", 0.1], 10],
    [["fgsm", 0.2], 10],
    [["fgsm", 0.3], 10],
    [["pgd", 0.1], 10],
    [["pgd", 0.2], 10],
    [["pgd", 0.3], 10],
]

# Construct eval_logs: list of dicts with evaluation results
# Each log has wandb_run_id, attack type, epsilon, and adv_acc
eval_logs = [
    {'wandb_run_id': 'run_123', 'attack': 'fgsm', 'epsilon': 0.1, 'adv_acc': 0.75},
    {'wandb_run_id': 'run_123', 'attack': 'fgsm', 'epsilon': 0.2, 'adv_acc': 0.70},
    {'wandb_run_id': 'run_123', 'attack': 'fgsm', 'epsilon': 0.3, 'adv_acc': 0.65},
    {'wandb_run_id': 'run_123', 'attack': 'pgd', 'epsilon': 0.1, 'adv_acc': 0.72},
    {'wandb_run_id': 'run_123', 'attack': 'pgd', 'epsilon': 0.2, 'adv_acc': 0.68},
    {'wandb_run_id': 'run_123', 'attack': 'pgd', 'epsilon': 0.3, 'adv_acc': 0.60},
]

# Construct defense: [defense_type, epsilon]
defense = ["madry", 0.3]

# Construct run_id
run_id = 'run_123'

ARGS = {
    'run_id': run_id,
    'atas': atas,
    'eval_logs': eval_logs,
    'defense': defense,
}
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 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
fog_creator Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/fog.py
code served (permissive licence) · get_code("6b64dfab81078b74")
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, 1, 4, 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 numpy as np

torch.manual_seed(0)

# fog_vars is used in wibbledmean function as fog_vars[var_num]
# It needs to be indexable and contain float values
# The function iterates through it, so we need enough elements
# Looking at the loop: stepsize starts at mapsize and halves each iteration
# For mapsize=4: stepsize goes 4, 2 (2 iterations)
# Each iteration calls fillsquares (var_num += 1) then filldiamonds (var_num += 2)
# So we need at least 3 elements per loop iteration, and 2 iterations = 6 elements minimum
fog_vars = np.array([0.5, 0.3, 0.2, 0.4, 0.1, 0.6], dtype=np.float32)

ARGS = {
    "fog_vars": fog_vars,
    "bsize": 2,
    "mapsize": 4,
    "wibbledecay": 1.75
}
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, 1, 4, 4) — shape (2, 1, 4, 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_attack Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_ata.py
code served (permissive licence) · get_code("ae65e44ca288d223")
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

torch.manual_seed(0)

log = {
    'attack': 'fgsm',
    'epsilon': 0.3,
    'n_iters': 10,
    'step_size': 0.01
}

ARGS = {
    'log': log
}
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
get_attack_types Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_uar.py
code served (permissive licence) · get_code("5b9cee62b4db0aa5")
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 dict. 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

torch.manual_seed(0)

# Analyze the function:
# - It takes 'atas' as a parameter
# - It iterates over 'atas'
# - For each 'ata', it accesses ata[0][0] to get the attack_type
# - It uses attack_type as a dictionary key
# - It appends 'ata' to lists in the dictionary
# So 'atas' should be a list of items where each item can be indexed [0][0]

# Create sample data structure
# ata should be indexable with [0][0], so it needs to be nested
# Let's create a list of items where each item is a list/tuple with at least 2 levels of nesting

atas = [
    [[1, 2, 3], [4, 5, 6]],      # attack_type = 1
    [[1, 7, 8], [9, 10, 11]],    # attack_type = 1
    [[2, 12, 13], [14, 15, 16]], # attack_type = 2
    [[2, 17, 18], [19, 20, 21]], # attack_type = 2
    [[3, 22, 23], [24, 25, 26]], # attack_type = 3
]

ARGS = {
    "atas": atas
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
get_attacks Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_ata.py
code served (permissive licence) · get_code("a5fe600c6f90a29e")
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 2, 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 numpy as np

torch.manual_seed(0)

# Construct valid arguments for get_attacks function
# The function expects all_logs: a list of log dictionaries
# Each log dict must have keys: 'attack', 'epsilon', 'n_iters', 'step_size'

all_logs = [
    {
        'attack': 'fgsm',
        'epsilon': 0.3,
        'n_iters': 1,
        'step_size': 0.3
    },
    {
        'attack': 'pgd',
        'epsilon': 0.1,
        'n_iters': 10,
        'step_size': 0.01
    },
    {
        'attack': 'fgsm',
        'epsilon': 0.3,
        'n_iters': 1,
        'step_size': 0.3
    },
    {
        'attack': 'cw',
        'epsilon': 0.2,
        'n_iters': 20,
        'step_size': 0.05
    }
]

ARGS = {
    'all_logs': all_logs
}
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 list, ndim 2, 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_defense_run_ids Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_uar.py
code served (permissive licence) · get_code("8cfbe3915f54e09f")
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

# Construct eval_logs with the structure expected by get_defense_run_ids
# The function iterates through eval_logs and accesses log['wandb_run_id']
eval_logs = [
    {'wandb_run_id': 'run_001', 'other_field': 'value1'},
    {'wandb_run_id': 'run_002', 'other_field': 'value2'},
    {'wandb_run_id': 'run_001', 'other_field': 'value3'},
    {'wandb_run_id': 'run_003', 'other_field': 'value4'},
]

ARGS = {
    'eval_logs': eval_logs
}
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_eps_params Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/attacks.py
code served (permissive licence) · get_code("d611a25ed152d734")
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 5, dtype family float; no output shape is stated. 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 and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    'base_eps': 0.1,
    'resol': 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 tuple, ndim 5, 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
get_gabor Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/gabor.py
code served (permissive licence) · get_code("2c107ae31d213cb3")
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 (16, 16) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - k_size: used in torch.linspace and meshgrid, should be an int
# - sigma: used in division, should be a float
# - Lambda: used in multiplication, should be a float
# - theta: used in torch.cos and torch.sin, should be a float or tensor

ARGS = {
    'k_size': 16,
    'sigma': 0.5,
    'Lambda': 0.1,
    'theta': torch.tensor(0.785)  # pi/4 radians
}
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 (16, 16) — shape (16, 16) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_imagenet_params Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/attacks.py
code served (permissive licence) · get_code("c6dc9d5916f95448")
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 5, dtype family float; no output shape is stated. 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 and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "resol": 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 tuple, ndim 5, 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
normalize Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/gabor.py
code served (permissive licence) · get_code("68e2adcd4a226d1f")
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, 4, 5) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small 4D tensor with batch_size=2
# Shape: (batch_size=2, channels=3, height=4, width=5)
orig = torch.randn(2, 3, 4, 5)

ARGS = {
    "orig": orig
}
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 4, dtype family float, shape (2, 3, 4, 5) — shape (2, 3, 4, 5) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_l2 Ran
ran, violated the contract we drafted
centerforaisafety/adversarial-corruptions/attacks/attacks.py
pointer only (licence: NONE) · get_code("c1e43d48e3be5a73")
How it ran

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

parse_logs Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/analysis/compute_ata.py
code served (permissive licence) · get_code("4195644f8fd630c7")
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 json
import tempfile
import os

# Create a temporary JSON file with test data
temp_dir = tempfile.mkdtemp()
log_file_list = os.path.join(temp_dir, "logs_list.json")
log_file_dict = os.path.join(temp_dir, "logs_dict.json")

# Write test data - list format
list_data = [
    {"epoch": 1, "loss": 0.5},
    {"epoch": 2, "loss": 0.3}
]
with open(log_file_list, 'w') as f:
    json.dump(list_data, f)

# Write test data - dict format
dict_data = {"epoch": 1, "loss": 0.5}
with open(log_file_dict, 'w') as f:
    json.dump(dict_data, f)

# Use the list format file for the test
ARGS = {
    "log_file": log_file_list
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
rgb_to_ycbcr_jpeg Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/jpeg.py
code served (permissive licence) · get_code("76d91377101614b7")
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. 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 numpy as np

torch.manual_seed(0)

# The function rgb_to_ycbcr_jpeg expects:
# - image: a torch tensor with shape that supports tensordot with a 3x3 matrix
# The image should have shape (..., 3) where the last dimension is RGB channels
# Let's create a small image tensor with shape (2, 4, 3) - batch of 2, spatial dim 4, 3 channels

image = torch.randn(2, 4, 3, dtype=torch.float32)

ARGS = {
    "image": image
}
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 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
tensor_clamp Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from joelsimonoff/gram-adversary-detection
pointer only; the stored copy get_code serves, from joelsimonoff/gram-adversary-detection, is pointer only (licence: NONE) · get_code("5b640b32a444ae38")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

tensor_clamp_l2 Ran
ran; our contract draft was wrong, not the code
centerforaisafety/adversarial-corruptions/attacks/attacks.py
pointer only (licence: NONE) · get_code("0cf0b2e3ac7eb3ab")
How it ran

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

valid_position Ran
ran on a synthesized input
ddkang/advex-uar/advex_uar/attacks/gabor.py
code served (permissive licence) · get_code("603ca4e51314b267")
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 bool, 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

torch.manual_seed(0)

# Analyze the function:
# - size: used in comparison (x >= size, y >= size), so it's a numeric value
# - x: used in comparison (x < 0, x >= size), so it's a numeric value
# - y: used in comparison (y < 0, y >= size), so it's a numeric value

# All parameters are simple integers/numbers used for boundary checking

ARGS = {
    "size": 10,
    "x": 5,
    "y": 7
}
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 bool, 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
rgb_to_ycbcr Not yet run
unverified (harvested, no recorded run)
ddkang/advex-uar/advex_uar/attacks/jpeg.py
code served (permissive licence) · get_code("f377568c1b1b92e8")
How it ran

no invocation recorded.

tensordot_pytorch Not yet run
unverified (harvested, no recorded run)
ddkang/advex-uar/advex_uar/attacks/jpeg.py
code served (permissive licence) · get_code("4b0cb77b372f4eef")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 later) and 16 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
tensor_clamp 4 other papers
3 dated earlier, 1 later
1508.06576 (2015-08)
A Neural Algorithm of Artistic Style
1901.09960 (2019-01)
Using Pre-Training Can Improve Model Robustness and Uncertainty
1906.12340 (2019-06)
Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty
and 1 more not named here
valid_position 1 other paper
1 dated earlier
1810.00470 (2018-10)
Procedural Noise Adversarial Examples for Black-Box Attacks on Deep Convolutional Networks

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

Adversarial robustness research primarily focuses on L_p perturbations, and most defenses are developed with identical training-time and test-time adversaries. However, in real-world applications developers are unlikely to have access to the full range of attacks or corruptions their system will face. Furthermore, worst-case inputs are likely to be diverse and need not be constrained to the L_p ball. To narrow in on this discrepancy between research and reality we introduce ImageNet-UA, a framework for evaluating model robustness against a range of unforeseen adversaries, including eighteen new non-L_p attacks. To perform well on ImageNet-UA, defenses must overcome a generalization gap and be robust to a diverse attacks not encountered during training. In extensive experiments, we find that existing robustness measures do not capture unforeseen robustness, that standard robustness techniques are beat by alternative training strategies, and that novel methods can improve unforeseen robustness. We present ImageNet-UA as a useful tool for the community for improving the worst-case behavior of machine learning systems.

For agents

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

get_harvested_code_for_paper("1908.08016")
get_code_for_paper("1908.08016")
have("1908.08016")

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

curl https://syntology.ai/api/ran/1908.08016.json

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

[![Syntology run record](https://syntology.ai/api/ran/1908.08016.svg)](https://syntology.ai/paper/1908.08016)

Connect an agent — have() is free.