SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2107.00644 · 2021

Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentation

arXiv · PDF · Open in the Atlas

Code that ran

17 of 20 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; 1 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 8 failures of our instrument; 3 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/2107.00644.json · badge: /api/ran/2107.00644.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
nicklashansen/dmcontrol-generalization-benchmark canonical (classifier label)
classifier basis: census
6 ran · 0 no recorded run
nicklashansen/svea-vit community (classifier label)
classifier basis: plan_row
9 ran · 3 no recorded run
SVJayanthi/dmcontrol-generalization-benchmark community (classifier label)
classifier basis: plan_row
2 ran · 0 no recorded run
FunctionStatusWhere it lives
batch_from_obs Ran
ran on a synthesized input
nicklashansen/dmcontrol-generalization-benchmark/src/augmentations.py
code served (permissive licence) · get_code("b25eb65554d61412")
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 (32, 4, 5, 6) 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)

# Test with a torch.Tensor observation (3D shape)
obs_torch = torch.randn(4, 5, 6)  # 3D tensor: channels=4, height=5, width=6

# Test with a numpy array observation (3D shape)
obs_numpy = np.random.randn(4, 5, 6)  # 3D array: channels=4, height=5, width=6

ARGS = {
    "obs": obs_torch,
    "batch_size": 32
}
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 (32, 4, 5, 6) — shape (32, 4, 5, 6) 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
cat Ran
ran; our contract draft was wrong, not the code
SVJayanthi/dmcontrol-generalization-benchmark/src/utils.py
code served (permissive licence) · get_code("6c2c84b4ccb7f17b")
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 (5, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create two small tensors with distinct shapes to probe dimensions
x = torch.randn(2, 4)
y = torch.randn(3, 4)

ARGS = {
    "x": x,
    "y": y,
    "axis": 0
}
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 (5, 4) — shape (5, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
do_green_screen Ran
ran on a synthesized input
nicklashansen/svea-vit/src/env/wrappers.py
code served (permissive licence) · get_code("09be0f6d9a709601")
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 3, dtype family int, shape (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 numpy as np
import torch
import torchvision.transforms.functional as TF

torch.manual_seed(0)

# Create a small RGB image with green screen (green background to be removed)
# Shape: (3, 4, 5) - 3 channels (RGB), 4 height, 5 width
x = np.zeros((3, 4, 5), dtype=np.uint8)
# Fill most with green (to be removed): G channel high, R and B low
x[0, :, :] = 100  # R channel
x[1, :, :] = 200  # G channel (green)
x[2, :, :] = 100  # B channel
# Add some non-green pixels in a corner
x[0, 0:2, 0:2] = 200  # R channel (red area)
x[1, 0:2, 0:2] = 50   # G channel (red area)
x[2, 0:2, 0:2] = 50   # B channel (red area)

# Create a background image with same shape
bg = np.ones((3, 4, 5), dtype=np.uint8) * 128
bg[0, :, :] = 50   # R channel (darker background)
bg[1, :, :] = 100  # G channel
bg[2, :, :] = 150  # B channel

ARGS = {
    'x': x,
    'bg': bg
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (3, 4, 5) — shape (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
drop_path Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
nicklashansen/svea-vit/src/algorithms/layers/drop.py
code served (permissive licence) · get_code("39eace7e2822504f")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 3, 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 a small tensor for testing
# Shape: (batch_size=2, channels=4, height=3, width=3) - 4D tensor
x = torch.randn(2, 4, 3, 3, dtype=torch.float32)

ARGS = {
    "x": x,
    "drop_prob": 0.5,
    "training": True,
}
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 4, dtype family float, shape (2, 4, 3, 3) — shape (2, 4, 3, 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
gaussian_logprob Ran
ran; our contract draft was wrong, not the code
nicklashansen/svea-vit/src/algorithms/modules.py
code served (permissive licence) · get_code("daa7b3a0355eb7c3")
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 Tensor, ndim 2, dtype family float, shape (2, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create a small noise tensor with shape (batch_size=2, feature_dim=4)
noise = torch.randn(2, 4)

# Create a small log_std tensor with shape (batch_size=2, feature_dim=4)
log_std = torch.randn(2, 4)

ARGS = {
    "noise": noise,
    "log_std": log_std,
}
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 2, dtype family float, shape (2, 1) — shape (2, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gaussian_logprob Ran
ran; our contract draft was wrong, not the code
nicklashansen/dmcontrol-generalization-benchmark/src/algorithms/modules.py
code served (permissive licence) · get_code("2df17488b927be0b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create small tensors with distinct dimensions
noise = torch.randn(2, 4)  # batch_size=2, dim=4
log_std = torch.randn(2, 4)  # batch_size=2, dim=4

ARGS = {
    "noise": noise,
    "log_std": log_std,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 1) — shape (2, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
listdir Ran
ran on a synthesized input
nicklashansen/svea-vit/src/utils.py
code served (permissive licence) · get_code("4cd7bffb72ec1909")
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 os
import tempfile

# Create a temporary directory with some test files
temp_dir = tempfile.mkdtemp()

# Create some test jpg files
test_files = ['image1.jpg', 'image2.jpg', 'photo.jpg']
for fname in test_files:
    fpath = os.path.join(temp_dir, fname)
    with open(fpath, 'w') as f:
        f.write('test')

# Create some non-jpg files to ensure filtering works
with open(os.path.join(temp_dir, 'document.txt'), 'w') as f:
    f.write('test')

ARGS = {
    'dir_path': temp_dir,
    'filetype': 'jpg',
    'sort': True
}
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
load_config Ran
ran on a synthesized input
nicklashansen/dmcontrol-generalization-benchmark/src/utils.py
code served (permissive licence) · get_code("5912a589c21e0e7f")
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 os
import json
import tempfile
import shutil

# Create a temporary directory structure for the test
temp_dir = tempfile.mkdtemp()
setup_dir = os.path.join(temp_dir, 'setup')
os.makedirs(setup_dir, exist_ok=True)

# Create a config file
config_data = {
    'model': 'bert',
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 10,
    'nested': {
        'value': 'test'
    }
}

config_path = os.path.join(setup_dir, 'config.cfg')
with open(config_path, 'w') as f:
    json.dump(config_data, f)

# Change to the temporary directory so relative path works
original_cwd = os.getcwd()
os.chdir(temp_dir)

# Define ARGS - the function accepts optional 'key' parameter
ARGS = {
    'key': None
}
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 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
make_dir Ran
ran; our contract draft was wrong, not the code
nicklashansen/svea-vit/src/utils.py
code served (permissive licence) · get_code("1ac4e5bffc3dbbd3")
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 str. 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 os
import tempfile

def make_dir(dir_path):
    try:
        os.makedirs(dir_path)
    except OSError:
        pass
    return dir_path

# Create a temporary directory path that can be safely created
temp_base = tempfile.gettempdir()
test_dir_path = os.path.join(temp_base, "test_dir_make_dir_setup")

ARGS = {
    "dir_path": test_dir_path
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
make_divisible Ran
ran, honoured the contract we drafted
nicklashansen/svea-vit/src/algorithms/layers/helpers.py
pointer only (licence: MIT); the stored copy get_code serves, from changlin31/DS-Net, is pointer only (licence: NONE) · get_code("5b2c72c30130d62a")
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 changlin31/DS-Net, 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.

random_conv Ran
ran on a synthesized input
nicklashansen/dmcontrol-generalization-benchmark/src/augmentations.py
code served (permissive licence) · get_code("0604211861b8f1e9")
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
import torch.nn.functional as F

torch.manual_seed(0)

# The function takes a single argument 'x' which is a tensor
# Looking at the function body:
# - x.shape is unpacked as (n, c, h, w) - a 4D tensor (batch, channels, height, width)
# - x[i:i+1] suggests batch indexing
# - x.device is used
# - x is reshaped and divided by 255.0
# - The output should have the same shape as input

# Create a small test tensor: batch=2, channels=3, height=4, width=5
x = torch.randint(0, 256, (2, 3, 4, 5), dtype=torch.float32)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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
random_conv Ran
ran on a synthesized input
nicklashansen/svea-vit/src/augmentations.py
code served (permissive licence) · get_code("e944e6e5e6f2648b")
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
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small input tensor with shape (batch=2, channels=3, height=4, width=5)
# The function expects x to have shape (n, c, h, w)
x = torch.randint(0, 256, (2, 3, 4, 5), dtype=torch.float32)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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
rgb_to_hsv Ran
ran on a synthesized input
nicklashansen/svea-vit/src/env/wrappers.py
code served (permissive licence) · get_code("3fa399e318655bf7")
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 numpy as np

torch.manual_seed(0)

# The function rgb_to_hsv takes three scalar values: r, g, b
# representing RGB color components (typically in range [0, 1] or [0, 255])
# The function performs comparisons and arithmetic operations on these values

# Valid test case: pure red color normalized to [0, 1]
r = 1.0
g = 0.0
b = 0.0

ARGS = {
    "r": r,
    "g": g,
    "b": b,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, 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
squash Ran
ran; our contract draft was wrong, not the code
nicklashansen/dmcontrol-generalization-benchmark/src/algorithms/modules.py
code served (permissive licence) · get_code("121cb336a1e1278a")
How it ran

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

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for testing
batch_size = 2
dim = 4

mu = torch.randn(batch_size, dim)
pi = torch.randn(batch_size, dim)
log_pi = torch.randn(batch_size, 1)

ARGS = {
    "mu": mu,
    "pi": pi,
    "log_pi": log_pi,
}
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 — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
squash Ran
ran; our contract draft was wrong, not the code
nicklashansen/svea-vit/src/algorithms/modules.py
code served (permissive licence) · get_code("adedb876e87c2947")
How it ran

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

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for the function parameters
# mu: input tensor that gets transformed by tanh
mu = torch.randn(2, 4)

# pi: policy output tensor that also gets transformed by tanh
pi = torch.randn(2, 4)

# log_pi: log probability tensor that gets adjusted based on the squashing transformation
log_pi = torch.randn(2, 1)

ARGS = {
    "mu": mu,
    "pi": pi,
    "log_pi": log_pi
}
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 — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
transform_batch_obs Ran
ran on a synthesized input
SVJayanthi/dmcontrol-generalization-benchmark/src/utils.py
code served (permissive licence) · get_code("c5bfdbe07ab3b06d")
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 4, dtype family int, shape (2, 9, 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
import numpy as np
import torchvision.transforms as transforms

torch.manual_seed(0)

# Create a small batch of observations
# Shape: (batch_size, channels, height, width)
# Using batch_size=2, channels=9 (3 frames * 3 channels each), height=4, width=3
batch_obs = np.random.randint(0, 256, size=(2, 9, 4, 3), dtype=np.uint8)

# Create a transform that operates on a single frame (3, height, width)
# The function calls transform(frame) where frame has shape (3, 4, 3)
transform = transforms.Compose([
    transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
])

ARGS = {
    "batch_obs": batch_obs,
    "transform": transform
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family int, shape (2, 9, 4, 3) — shape (2, 9, 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
trunc_normal_ Ran
ran; our contract draft was wrong, not the code
nicklashansen/dmcontrol-generalization-benchmark/src/algorithms/modules.py
code served (permissive licence) · get_code("1387f0bafec9b149")
How it ran

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

import torch
import math

torch.manual_seed(0)

# Create a small tensor to be filled with truncated normal values
tensor = torch.empty(2, 4)

# Define parameters for the truncated normal distribution
mean = 0.0
std = 1.0
a = -2.0
b = 2.0

ARGS = {
    "tensor": tensor,
    "mean": mean,
    "std": std,
    "a": a,
    "b": b
}
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
drop_block_2d Not yet run
unverified (harvested, no recorded run)
nicklashansen/svea-vit/src/algorithms/layers/drop.py
code served (permissive licence) · get_code("653d5a3aa2c473be")
How it ran

no invocation recorded.

drop_block_fast_2d Not yet run
unverified (harvested, no recorded run)
nicklashansen/svea-vit/src/algorithms/layers/drop.py
pointer only (licence: MIT); the stored copy get_code serves, from summitgao/ss-mae, is pointer only (licence: NONE) · get_code("97aa7c942cfb7d3d")
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 summitgao/ss-mae, 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. No invocation recorded.

vit_crop Not yet run
unverified (harvested, no recorded run)
nicklashansen/svea-vit/src/augmentations.py
code served (permissive licence) · get_code("847b6ac61bd69609")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 11 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 11 later) and 6 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
drop_path 224 other papers
8 dated earlier, 215 later, 1 undated · 38 only as an assembly brick
2003.12039 (2020-03)
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
2007.09384 (2020-07)
Multi-Scale Positive Sample Refinement for Few-Shot Object Detection
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
and 221 more not named here
gaussian_logprob 15 other papers
9 dated earlier, 6 later · 3 only as an assembly brick
1910.01741 (2019-10)
Improving Sample Efficiency in Model-Free Reinforcement Learning from Images
2004.04136 (2020-04)
CURL: Contrastive Unsupervised Representations for Reinforcement Learning
2004.14990 (2020-04)
Reinforcement Learning with Augmented Data
and 12 more not named here
squash 14 other papers
8 dated earlier, 6 later · 3 only as an assembly brick
1910.01741 (2019-10)
Improving Sample Efficiency in Model-Free Reinforcement Learning from Images
2004.04136 (2020-04)
CURL: Contrastive Unsupervised Representations for Reinforcement Learning
2004.14990 (2020-04)
Reinforcement Learning with Augmented Data
and 11 more not named here
drop_block_2d 12 other papers
1 dated earlier, 11 later
2106.00666 (2021-06)
You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection
2108.05045 (2021-08)
Semi-Supervised Domain Generalizable Person Re-Identification
2110.07719 (2021-10)
Certified Patch Robustness via Smoothed Vision Transformers
and 9 more not named here
drop_block_fast_2d 11 other papers
1 dated earlier, 10 later
2106.00666 (2021-06)
You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection
2108.05045 (2021-08)
Semi-Supervised Domain Generalizable Person Re-Identification
2110.07719 (2021-10)
Certified Patch Robustness via Smoothed Vision Transformers
and 8 more not named here
make_dir 7 other papers
7 later
2210.07241 (2022-10)
Visual Reinforcement Learning with Self-Supervised 3D Representations
2212.05698 (2022-12)
MoDem: Accelerating Visual Model-Based Reinforcement Learning with Demonstrations
2312.15339 (2023-12)
MaDi: Learning to Mask Distractions for Generalization in Visual Deep Reinforcement Learning
and 4 more not named here
make_divisible 5 other papers
2 dated earlier, 3 later
1702.01105 (2017-02)
Joint 2D-3D-Semantic Data for Indoor Scene Understanding
2103.13258 (2021-03)
Dynamic Slimmable Network
Guo_Improving_Robustness_of_Vision_Transformers_by_Reducing_Sensitivity_To_Patch_CVPR_2023_paper (2023 (year only))
Improving Robustness of Vision Transformers by Reducing Sensitivity to Patch Corruptions
and 2 more not named here
cat 2 other papers
2 later
2210.07241 (2022-10)
Visual Reinforcement Learning with Self-Supervised 3D Representations
2312.15339 (2023-12)
MaDi: Learning to Mask Distractions for Generalization in Visual Deep Reinforcement Learning
load_config 1 other paper
1 later
2210.07241 (2022-10)
Visual Reinforcement Learning with Self-Supervised 3D Representations
random_conv 1 other paper
1 later
2210.07241 (2022-10)
Visual Reinforcement Learning with Self-Supervised 3D Representations
trunc_normal_ 1 other paper
1 later
2312.15339 (2023-12)
MaDi: Learning to Mask Distractions for Generalization in Visual Deep Reinforcement 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

While agents trained by Reinforcement Learning (RL) can solve increasingly challenging tasks directly from visual observations, generalizing learned skills to novel environments remains very challenging. Extensive use of data augmentation is a promising technique for improving generalization in RL, but it is often found to decrease sample efficiency and can even lead to divergence. In this paper, we investigate causes of instability when using data augmentation in common off-policy RL algorithms. We identify two problems, both rooted in high-variance Q-targets. Based on our findings, we propose a simple yet effective technique for stabilizing this class of algorithms under augmentation. We perform extensive empirical evaluation of image-based RL using both ConvNets and Vision Transformers (ViT) on a family of benchmarks based on DeepMind Control Suite, as well as in robotic manipulation tasks. Our method greatly improves stability and sample efficiency of ConvNets under augmentation, and achieves generalization results competitive with state-of-the-art methods for image-based RL in environments with unseen visuals. We further show that our method scales to RL with ViT-based architectures, and that data augmentation may be especially important in this setting.

For agents

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

get_harvested_code_for_paper("2107.00644")
get_code_for_paper("2107.00644")
have("2107.00644")

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

curl https://syntology.ai/api/ran/2107.00644.json

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

[![Syntology run record](https://syntology.ai/api/ran/2107.00644.svg)](https://syntology.ai/paper/2107.00644)

Connect an agent — have() is free.