SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2301.05158 · ICLR · 2023

SemPPL: Predicting pseudo-labels for better contrastive representations

Charles Blundell, Lars Buesing, Felix Hill, Razvan Pascanu, Jovana Mitrovic, Florian Strub, Pierre Richemond, Nenad Tomasev, Matko Bošnjak, Jacob Walker

arXiv · PDF · Open in the Atlas

Code that ran

12 of 16 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, 2 violated it and 10 have no contract verdict; 0 failures of our instrument; 4 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2301.05158.json · badge: /api/ran/2301.05158.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
google-deepmind/semppl canonical (classifier label)
classifier basis not recorded on the link
12 ran · 2 no recorded run
deepmind/semppl alias (classifier label)
classifier basis: classifier_2025
0 ran · 2 no recorded run
FunctionStatusWhere it lives
exclude_bias_and_norm Ran
ran on a synthesized input
google-deepmind/semppl/utils/optimizers.py
code served (permissive licence) · get_code("7cf26db115ac0335")
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 jax.numpy as jnp
from typing import Tuple, Any

# Build test arguments for exclude_bias_and_norm function
# The function takes:
# - path: Tuple[Any] - a tuple representing a path through a nested structure
# - val: jnp.ndarray - a JAX numpy array (though it's not used in the function)

# Create a small JAX array
test_array = jnp.array([1.0, 2.0, 3.0, 4.0])

# Test case 1: path that should return False (contains 'b' at end)
path_with_bias = ("layer", "dense", "b")

# Test case 2: path that should return False (contains 'norm' in penultimate position)
path_with_norm = ("layer", "norm", "weight")

# Test case 3: path that should return True (no bias or norm)
path_normal = ("layer", "dense", "w")

# Use the third case as the main test
ARGS = {
    "path": path_normal,
    "val": test_array
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — 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
hsv_to_rgb Ran
ran, violated the contract we drafted
google-deepmind/semppl/utils/augmentations.py
code served (permissive licence) · get_code("56b9317b9ed66304")
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 3, 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 jax.numpy as jnp

# Create small JAX arrays for h, s, v parameters
# All should be float arrays with values in range [0, 1]
h = jnp.array([[0.0, 0.2, 0.5, 0.8],
               [0.1, 0.3, 0.7, 0.9]], dtype=jnp.float32)

s = jnp.array([[0.5, 0.6, 0.7, 0.8],
               [0.4, 0.5, 0.6, 0.7]], dtype=jnp.float32)

v = jnp.array([[0.8, 0.9, 0.7, 0.6],
               [0.9, 0.8, 0.5, 0.4]], dtype=jnp.float32)

ARGS = {
    "h": h,
    "s": s,
    "v": v,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, 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
l2_normalize Ran
ran on a synthesized input
google-deepmind/semppl/utils/helpers.py
code served (permissive licence) · get_code("8df258170fa3cb79")
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 ArrayImpl, 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 jax
import jax.numpy as jnp
import numpy as np

# Set seed for reproducibility
np.random.seed(0)

# Create a small test array with shape (2, 4)
x = jnp.array(np.random.randn(2, 4), dtype=jnp.float32)

# Define arguments for l2_normalize
ARGS = {
    "x": x,
    "axis": None,
    "epsilon": 1e-12
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, 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
learning_schedule Ran
ran on a synthesized input
google-deepmind/semppl/utils/schedules.py
code served (permissive licence) · get_code("5e9e28416e760f15")
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 ArrayImpl, 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 jax.numpy as jnp

# Set up arguments for learning_schedule function
global_step = jnp.array(100, dtype=jnp.int32)  # Current training step
batch_size = 32  # Batch size for training
base_learning_rate = 0.1  # Base learning rate
total_steps = 1000  # Total number of training steps
warmup_steps = 100  # Number of warmup steps

ARGS = {
    "global_step": global_step,
    "batch_size": batch_size,
    "base_learning_rate": base_learning_rate,
    "total_steps": total_steps,
    "warmup_steps": warmup_steps,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, 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
load_checkpoint Ran
ran on a synthesized input
google-deepmind/semppl/utils/checkpointing.py
code served (permissive licence) · get_code("5b529484c13fcd15")
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 tempfile
import os
import dill
import torch

# Create a temporary checkpoint file with valid checkpoint data
checkpoint_data = {
    'step': 42,
    'model_weights': torch.randn(4, 3),
    'optimizer_state': {'lr': 0.001}
}

# Create temporary directory and write checkpoint
temp_dir = tempfile.mkdtemp()
checkpoint_path = os.path.join(temp_dir, 'checkpoint.pkl')

with open(checkpoint_path, 'wb') as f:
    dill.dump(checkpoint_data, f)

ARGS = {
    'checkpoint_path': checkpoint_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
absl, dill, 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
normalize_images Ran
ran on a synthesized input
google-deepmind/semppl/utils/dataset.py
code served (permissive licence) · get_code("9299bed7da61102a")
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 ArrayImpl, ndim 4, dtype family float, shape (2, 4, 5, 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 jax.numpy as jnp

# Create a small batch of images
# Shape: (batch_size, height, width, channels)
# Using batch=2, height=4, width=5, channels=3
batch_size = 2
height = 4
width = 5
channels = 3

# Create random image data in typical [0, 1] or [0, 255] range
# Using [0, 1] range as is common for normalized inputs
images = jnp.ones((batch_size, height, width, channels)) * 0.5

ARGS = {
    "images": images
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 4, dtype family float, shape (2, 4, 5, 3) — shape (2, 4, 5, 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
rgb_to_hsv Ran
ran, violated the contract we drafted
google-deepmind/semppl/utils/augmentations.py
code served (permissive licence) · get_code("bf9bb14716c6dab9")
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 3, 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 jax.numpy as jnp

# Create small sample tensors for R, G, B color components
# Values should be in range [0, 1] for proper color conversion
# Using distinct shapes to avoid ambiguity: batch of 2, with 4 color samples each
r = jnp.array([[0.5, 0.2, 0.8, 1.0],
                [0.0, 0.3, 0.6, 0.9]])

g = jnp.array([[0.5, 0.4, 0.6, 0.5],
                [0.2, 0.5, 0.7, 0.8]])

b = jnp.array([[0.5, 0.6, 0.4, 0.2],
                [0.4, 0.7, 0.5, 0.6]])

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
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, 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
scale_by_lars Ran
ran on a synthesized input
google-deepmind/semppl/utils/optimizers.py
code served (permissive licence) · get_code("79415fc4013fff5c")
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 GradientTransformation. 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 jax
import jax.numpy as jnp
import optax
from typing import Any, Tuple

# Create small parameter structure
params = {
    'layer1': jnp.ones((4, 3)),
    'layer2': jnp.ones((3, 2)),
}

# Create a simple filter function that returns True for some parameters
def simple_filter_fn(path: Tuple[Any], param: jnp.ndarray) -> jnp.ndarray:
    # Filter function that selects certain parameters
    # Returns True (1.0) for 'layer1', False (0.0) for 'layer2'
    if len(path) > 0 and 'layer1' in str(path[0]):
        return True
    return False

ARGS = {
    'momentum': 0.9,
    'eta': 0.001,
    'filter_fn': simple_filter_fn,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, optax, tree — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GradientTransformation — a GradientTransformation output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a GradientTransformation output the recorded ndim 1 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
softmax_cross_entropy Ran
ran on a synthesized input
google-deepmind/semppl/utils/helpers.py
code served (permissive licence) · get_code("37ae8f08fc24a7d8")
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 ArrayImpl, 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 jax
import jax.numpy as jnp

# Set seed for reproducibility
jax.config.update('jax_default_prng_impl', 'threefry2x32')
key = jax.random.PRNGKey(0)

# Create small test tensors
batch_size = 2
num_classes = 4

# Logits: shape (batch_size, num_classes)
logits = jnp.array([
    [1.0, 2.0, 0.5, 1.5],
    [0.2, 0.8, 1.2, 0.5]
], dtype=jnp.float32)

# One-hot encoded labels: shape (batch_size, num_classes)
labels = jnp.array([
    [0.0, 1.0, 0.0, 0.0],
    [1.0, 0.0, 0.0, 0.0]
], dtype=jnp.float32)

# Reduction mode
reduction = 'mean'

ARGS = {
    'logits': logits,
    'labels': labels,
    'reduction': reduction
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, 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
target_ema Ran
ran on a synthesized input
google-deepmind/semppl/utils/schedules.py
code served (permissive licence) · get_code("8c0e47e0bfa8f06f")
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 ArrayImpl, 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 jax.numpy as jnp

# Create arguments for target_ema function
global_step = jnp.array(50)
base_ema = 0.999
max_steps = 1000

ARGS = {
    "global_step": global_step,
    "base_ema": base_ema,
    "max_steps": max_steps,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, 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
topk_accuracy Ran
ran on a synthesized input
google-deepmind/semppl/utils/helpers.py
code served (permissive licence) · get_code("73ed5cd901b2ff52")
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 ArrayImpl, ndim 1, dtype family float, shape (2,) 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 jax.numpy as jnp

# Create sample logits and labels
batch_size = 2
num_classes = 4
topk_value = 2

# logits: shape (batch_size, num_classes) - raw model outputs
logits = jnp.array([
    [0.1, 0.9, 0.5, 0.2],
    [0.8, 0.1, 0.3, 0.6]
], dtype=jnp.float32)

# labels: shape (batch_size,) - ground truth class indices
labels = jnp.array([1, 0], dtype=jnp.int32)

# topk: number of top predictions to consider
topk = topk_value

# ignore_label_above: optional threshold for filtering labels
ignore_label_above = None

ARGS = {
    'logits': logits,
    'labels': labels,
    'topk': topk,
    'ignore_label_above': ignore_label_above
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (2,) — shape (2,) 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
transpose_images Ran
ran on a synthesized input
google-deepmind/semppl/utils/dataset.py
code served (permissive licence) · get_code("301ebdfc332bbcc9")
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 jax.numpy as jnp
import numpy as np

# Create a batch with 'images' key
# The function transposes from (batch, height, width, channels) to (channels, batch, height, width)
# So we need shape (batch_size, height, width, channels) = (2, 3, 4, 5)
batch_with_images = {
    'images': np.random.randn(2, 3, 4, 5).astype(np.float32),
    'labels': np.array([0, 1])
}

ARGS = {
    'batch': batch_with_images
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — 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
ResNetTorso Not yet run
unverified (harvested, no recorded run)
deepmind/semppl/utils/networks.py
code served (permissive licence) · get_code("ec03c773fc93b2dd")
How it ran

no invocation recorded.

add_weight_decay Not yet run
unverified (harvested, no recorded run)
google-deepmind/semppl/utils/optimizers.py
code served (permissive licence) · get_code("43422aa37e912a67")
How it ran

no invocation recorded.

check_length Not yet run
unverified (harvested, no recorded run)
deepmind/semppl/utils/networks.py
code served (permissive licence) · get_code("05611d15ac79d977")
How it ran

no invocation recorded.

postprocess Not yet run
unverified (harvested, no recorded run)
google-deepmind/semppl/utils/augmentations.py
code served (permissive licence) · get_code("c14a511164946e31")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 14 are also present in other papers whose code we hold (14 with at least one paper dated earlier, 0 the same month, 3 later) and 2 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
hsv_to_rgb 2 other papers
1 dated earlier, 1 later
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
2307.15157 (2023-07)
R-LPIPS: An Adversarially Robust Perceptual Similarity Metric
postprocess 2 other papers
1 dated earlier, 1 later
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
2307.15157 (2023-07)
R-LPIPS: An Adversarially Robust Perceptual Similarity Metric
rgb_to_hsv 2 other papers
1 dated earlier, 1 later
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
2307.15157 (2023-07)
R-LPIPS: An Adversarially Robust Perceptual Similarity Metric
add_weight_decay 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
exclude_bias_and_norm 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
l2_normalize 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
learning_schedule 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
load_checkpoint 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
normalize_images 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
scale_by_lars 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
softmax_cross_entropy 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
target_ema 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
topk_accuracy 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
transpose_images 1 other paper
1 dated earlier
2201.05119 (2022-01)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?

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

Learning from large amounts of unsupervised data and a small amount of supervision is an important open problem in computer vision. We propose a new semisupervised learning method, Semantic Positives via Pseudo-Labels (SEMPPL), that combines labelled and unlabelled data to learn informative representations. Our method extends self-supervised contrastive learning-where representations are shaped by distinguishing whether two samples represent the same underlying datum (positives) or not (negatives)-with a novel approach to selecting positives. To enrich the set of positives, we leverage the few existing ground-truth labels to predict the missing ones through a k-nearest neighbours classifier by using the learned embeddings of the labelled data. We thus extend the set of positives with datapoints having the same pseudo-label and call these semantic positives. We jointly learn the representation and predict bootstrapped pseudolabels. This creates a reinforcing cycle. Strong initial representations enable better pseudo-label predictions which then improve the selection of semantic positives and lead to even better representations. SEMPPL outperforms competing semisupervised methods setting new state-of-the-art performance of 68.5% and 76% top-1 accuracy when using a ResNet-50 and training on 1% and 10% of labels on ImageNet, respectively. Furthermore, when using selective kernels, SEMPPL significantly outperforms previous state-of-the-art achieving 72.3% and 78.3% top-1 accuracy on ImageNet with 1% and 10% labels, respectively, which improves absolute +7.8% and +6.2% over previous work. SEMPPL also exhibits stateof-the-art performance over larger ResNet models as well as strong robustness, out-of-distribution and transfer performance. We release the checkpoints and the evaluation code at https://github.com/deepmind/semppl.

For agents

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

get_harvested_code_for_paper("2301.05158")
get_code_for_paper("2301.05158")
have("2301.05158")

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

curl https://syntology.ai/api/ran/2301.05158.json

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

[![Syntology run record](https://syntology.ai/api/ran/2301.05158.svg)](https://syntology.ai/paper/2301.05158)

Connect an agent — have() is free.