SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.16241 · 2020

The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization

Dan Hendrycks, Steven Basart, Jacob Steinhardt, Dawn Song, Saurav Kadavath, Tyler Zhu, Samyak Parajuli, Justin Gilmer, Norman Mu, Frank Wang, Evan Dorundo, Rahul Desai, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

14 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 12 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/2006.16241.json · badge: /api/ran/2006.16241.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
hendrycks/imagenet-r marked official in Papers with Code
Papers with Code archive
14 ran · 0 no recorded run
FunctionStatusWhere it lives
AugMix Ran
ran on a synthesized input · constructed an object; no result was computed
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("90f0b6606b7bec55")
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 an instance of its own class, AugMix: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from PIL import Image

# Create a minimal dataset that AugMix wraps
class MinimalDataset(torch.utils.data.Dataset):
    def __init__(self, num_samples=2):
        self.num_samples = num_samples
    
    def __getitem__(self, idx):
        # Return a PIL image and a label
        pil_img = Image.new('RGB', (256, 256), color=(73, 109, 137))
        label = idx % 2
        return pil_img, label
    
    def __len__(self):
        return self.num_samples

# Create the base dataset
base_dataset = MinimalDataset(num_samples=2)

# Define ARGS dictionary with all parameters for AugMix
ARGS = {
    'dataset': base_dataset,
    'severity': 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type AugMix — the run constructed an instance of its own class and did not call it; no shape 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
autocontrast Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("48502588e268dd8f")
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 Image. 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.

from PIL import Image, ImageOps
import numpy as np

# Create a small PIL Image for testing
img_array = np.random.randint(0, 256, (10, 10, 3), dtype=np.uint8)
pil_img = Image.fromarray(img_array, mode='RGB')

ARGS = {
    'pil_img': pil_img,
    'level': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
brightness Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("7fa83d532dab3fa4")
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 Image. 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
from PIL import Image, ImageEnhance

# Create a small PIL Image for testing
pil_img = Image.new('RGB', (4, 3), color=(100, 150, 200))

# Define ARGS mapping all parameters of the brightness function
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
color Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("f53b32f9516fdaa1")
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 Image. 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
from PIL import Image, ImageEnhance

# Create a small PIL image for testing
pil_img = Image.new('RGB', (4, 3), color=(100, 150, 200))

# Define ARGS dictionary with all parameters for the color function
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
contrast Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("6e50ec13656740c3")
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 Image. 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
from PIL import Image, ImageEnhance

# Create a small PIL image for testing
pil_img = Image.new('RGB', (4, 3), color=(100, 150, 200))

# Define ARGS with all parameters for the contrast function
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
equalize Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("cbb8d1e8526fa7eb")
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 Image. 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.

from PIL import Image, ImageOps
import numpy as np

# Create a small PIL Image for testing
# Create a simple RGB image (4x3 pixels)
img_array = np.array([
    [[255, 0, 0], [0, 255, 0], [0, 0, 255], [128, 128, 128]],
    [[100, 100, 100], [200, 200, 200], [50, 50, 50], [150, 150, 150]],
    [[10, 20, 30], [40, 50, 60], [70, 80, 90], [110, 120, 130]],
    [[230, 240, 250], [20, 30, 40], [100, 110, 120], [200, 210, 220]]
], dtype=np.uint8)

pil_img = Image.fromarray(img_array, mode='RGB')

ARGS = {
    'pil_img': pil_img,
    'level': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
posterize Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("e51e7aefc03823aa")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small test image (PIL Image object)
test_image = Image.new('RGB', (16, 16), color=(100, 150, 200))

# Define ARGS dict with all parameters for the posterize function
ARGS = {
    'pil_img': test_image,
    'level': 6
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
rotate Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("f3de03c4a896f7e5")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small PIL Image object
# The function expects a PIL Image that has a rotate method
pil_img = Image.new('RGB', (32, 32), color='red')

# Define ARGS with all parameters for the rotate function
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
sharpness Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("c55e7427481ee18e")
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 Image. 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
from PIL import Image, ImageEnhance

# Create a small PIL Image object
# Create a simple RGB image (2x2 pixels)
pil_img = Image.new('RGB', (2, 2), color=(100, 150, 200))

# Define arguments
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
shear_x Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("8b301c76d5b96b5f")
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 Image. 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
from PIL import Image

# Create a small PIL image (256x256 as required by the function)
pil_img = Image.new('RGB', (256, 256), color='red')

# Define the level parameter - should be a numeric value
level = 5

# Create the ARGS dictionary with all parameters
ARGS = {
    'pil_img': pil_img,
    'level': level
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
shear_y Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("74de255c18b450ef")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small PIL image for testing
pil_img = Image.new('RGB', (256, 256), color='red')

ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
solarize Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("ea3db097ab0c59d3")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small PIL image for testing
pil_img = Image.new('RGB', (4, 3), color='red')

# Define level parameter - should be a number between 0 and some positive value
level = 5

# Construct ARGS dict with all parameters
ARGS = {
    'pil_img': pil_img,
    'level': level
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
translate_x Ran
ran; our contract draft was wrong, not the code
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("68ee08ee193d291c")
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 Image. 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.

from PIL import Image, ImageOps, ImageEnhance
import numpy as np

# Create a small PIL image for testing
pil_img = Image.new('RGB', (256, 256), color=(128, 128, 128))

# Define level parameter - an integer that will be used in the function
level = 5

# Construct ARGS dictionary with all parameters for translate_x
ARGS = {
    'pil_img': pil_img,
    'level': level
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
translate_y Ran
ran on a synthesized input
hendrycks/imagenet-r/deepfashion_remix/augmix.py
code served (permissive licence) · get_code("49792e8ad13a4ec5")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small PIL image for testing
# The function expects a PIL Image object
pil_img = Image.new('RGB', (256, 256), color='red')

# Define ARGS dict with all parameters for translate_y function
ARGS = {
    'pil_img': pil_img,
    'level': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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

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 (0 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
autocontrast
brick of this paper's assembly: data_transform
4 other papers
4 later
2110.09506 (2021-10)
MEMO: Test Time Robustness via Adaptation and Augmentation
2310.20199 (2023-10)
In Search of Lost Online Test-time Adaptation: A Survey
2401.08328 (2024-01)
Un-Mixing Test-Time Normalization Statistics: Combatting Label Temporal Correlation
and 1 more not named here
equalize
brick of this paper's assembly: data_transform
4 other papers
4 later
2110.09506 (2021-10)
MEMO: Test Time Robustness via Adaptation and Augmentation
2310.20199 (2023-10)
In Search of Lost Online Test-time Adaptation: A Survey
2401.08328 (2024-01)
Un-Mixing Test-Time Normalization Statistics: Combatting Label Temporal Correlation
and 1 more not named here
rotate
brick of this paper's assembly: data_transform
4 other papers
4 later
2110.09506 (2021-10)
MEMO: Test Time Robustness via Adaptation and Augmentation
2310.20199 (2023-10)
In Search of Lost Online Test-time Adaptation: A Survey
2401.08328 (2024-01)
Un-Mixing Test-Time Normalization Statistics: Combatting Label Temporal Correlation
and 1 more not named here
posterize
brick of this paper's assembly: data_transform
1 other paper
1 later
2110.09506 (2021-10)
MEMO: Test Time Robustness via Adaptation and Augmentation
solarize
brick of this paper's assembly: data_transform
1 other paper
1 later
2110.09506 (2021-10)
MEMO: Test Time Robustness via Adaptation and Augmentation

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 introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on realworld distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pretrained with 1000× more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness. * Equal contribution. 1 UC Berkeley, 2 UChicago, 3 Google. Code is available at https://github.com/hendrycks/imagenet-r.

For agents

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

get_harvested_code_for_paper("2006.16241")
get_code_for_paper("2006.16241")
have("2006.16241")

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

curl https://syntology.ai/api/ran/2006.16241.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.16241.svg)](https://syntology.ai/paper/2006.16241)

Connect an agent — have() is free.