SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.10029 · NeurIPS · 2020

Big Self-Supervised Models are Strong Semi-Supervised Learners

Ting Chen, Simon Kornblith, Mohammad Norouzi, Geoffrey Hinton, Kevin Swersky

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 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, 0 violated it and 4 have no contract verdict; 0 failures of our instrument; 2 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/2006.10029.json · badge: /api/ran/2006.10029.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
serre-lab/prj_selfsup community (classifier label)
classifier basis not recorded on the link
4 ran · 2 no recorded run
FunctionStatusWhere it lives
attractive_loss Ran
ran on a synthesized input
serre-lab/prj_selfsup/model/objective.py
code served (permissive licence) · get_code("a7a5dcdaf8aafb54")
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 tensorflow.compat.v1 as tf

tf.disable_v2_behavior()

# Create a small hidden tensor of shape (batch_size, dim)
# The function splits this into two equal parts, so batch_size must be even
batch_size = 4
dim = 8

hidden = tf.constant([
    [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8],
    [0.15, 0.25, 0.35, 0.45, 0.55, 0.65, 0.75, 0.85],
    [0.12, 0.22, 0.32, 0.42, 0.52, 0.62, 0.72, 0.82],
    [0.18, 0.28, 0.38, 0.48, 0.58, 0.68, 0.78, 0.88],
], dtype=tf.float32)

ARGS = {
    'hidden': hidden,
    'temperature': 1.0,
    'hidden_norm': True,
    'weights': 1.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — 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
attractive_repulsive_loss Ran
ran on a synthesized input
serre-lab/prj_selfsup/model/objective.py
code served (permissive licence) · get_code("79e0efae44111eea")
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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create a small hidden tensor: shape (4, 8) - batch size 4, embedding dim 8
# This will be split into hidden1 and hidden2, so total batch is 4
hidden_np = np.random.randn(4, 8).astype(np.float32)
hidden = tf.constant(hidden_np)

# Simple scalar weight
weights = 1.0

# Temperature parameter
temperature = 0.5

# hidden_norm flag
hidden_norm = True

# tpu_context as None (CPU execution)
tpu_context = None

ARGS = {
    "hidden": hidden,
    "hidden_norm": hidden_norm,
    "temperature": temperature,
    "tpu_context": tpu_context,
    "weights": weights,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — 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
resnet_autoencoder_v1_generator Ran
ran on a synthesized input
serre-lab/prj_selfsup/model/resnet_ae.py
code served (permissive licence) · get_code("ac78dbf80b2d832a")
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 function. 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 tensorflow.compat.v1 as tf
import numpy as np
from absl import flags

# Disable v2 behavior for TensorFlow v1
tf.disable_v2_behavior()

FLAGS = flags.FLAGS
flags.DEFINE_bool('use_td_loss', True, 'Use temporal difference loss')

# Create minimal encoder function
def minimal_encoder(inputs, is_training):
    """Minimal encoder that returns features and block_activities."""
    # inputs shape: [B, H, W, C]
    # Return features [B, H, W, C_feat] and block_activities (can be empty dict)
    with tf.variable_scope('encoder_impl', reuse=tf.AUTO_REUSE):
        features = tf.layers.conv2d(
            inputs, filters=8, kernel_size=3, padding='same', 
            activation=tf.nn.relu, name='conv1'
        )
    block_activities = {}
    return features, block_activities

# Create minimal decoder function
def minimal_decoder(features, block_activities, is_training, skip):
    """Minimal decoder that reconstructs images."""
    # features shape: [B, H, W, C_feat + A]
    # Return reconstructed images [B, H, W, 3]
    with tf.variable_scope('decoder_impl', reuse=tf.AUTO_REUSE):
        recon = tf.layers.conv2d(
            features, filters=3, kernel_size=3, padding='same',
            activation=None, name='conv_out'
        )
    return recon

# Create minimal metric function
def minimal_metric(images, is_training):
    """Minimal metric that returns embeddings."""
    # images shape: [B, H, W, 3]
    # Return embeddings [B, embedding_dim]
    with tf.variable_scope('metric_impl', reuse=tf.AUTO_REUSE):
        # Global average pool
        pooled = tf.reduce_mean(images, axis=[1, 2])  # [B, 3]
        # Dense layer
        embeddings = tf.layers.dense(pooled, units=6, name='embedding')
    return embeddings

ARGS = {
    'encoder': minimal_encoder,
    'decoder': minimal_decoder,
    'metric': minimal_metric,
    'skip': True,
    'mask_augs': 0.5,
    'greyscale_viz': False,
    'data_format': 'channels_last',
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
absl, numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
supervised_loss Ran
ran on a synthesized input
serre-lab/prj_selfsup/model/objective.py
code served (permissive licence) · get_code("e888c30f5b90e24a")
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 SymbolicTensor. 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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create small valid tensors for softmax cross entropy
batch_size = 2
num_classes = 4

# labels: one-hot encoded, shape [batch_size, num_classes]
labels = tf.constant([
    [1.0, 0.0, 0.0, 0.0],
    [0.0, 1.0, 0.0, 0.0]
], dtype=tf.float32)

# logits: shape [batch_size, num_classes]
logits = tf.constant([
    [2.0, 1.0, 0.5, 0.1],
    [0.1, 2.5, 1.0, 0.5]
], dtype=tf.float32)

# weights: shape [batch_size] - weight for each example
weights = tf.constant([1.0, 1.0], dtype=tf.float32)

ARGS = {
    'labels': labels,
    'logits': logits,
    'weights': weights
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor 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
get_train_steps Not yet run
unverified (harvested, no recorded run)
serre-lab/prj_selfsup/model/model_util.py
code served (permissive licence) · get_code("14ec8ff095316493")
How it ran

no invocation recorded.

learning_rate_schedule Not yet run
unverified (harvested, no recorded run)
serre-lab/prj_selfsup/model/model_util.py
code served (permissive licence) · get_code("ea360af54446b02b")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 1 the same month, 2 later) and 4 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
get_train_steps 4 other papers
1 the same month, 3 later
2006.05576 (2020-06)
Self-supervised Learning from a Multi-view Perspective
2109.12909 (2021-09)
Compressive Visual Representations
aaai_20650 (2022 (year only))
Adversarial Examples Can Be Effective Data Augmentation for Unsupervised Machine Learning
and 1 more not named here
learning_rate_schedule 1 other paper
1 later
aaai_20650 (2022 (year only))
Adversarial Examples Can Be Effective Data Augmentation for Unsupervised Machine 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

One paradigm for learning from few labeled examples while making best use of a large amount of unlabeled data is unsupervised pretraining followed by supervised fine-tuning. Although this paradigm uses unlabeled data in a task-agnostic way, in contrast to common approaches to semi-supervised learning for computer vision, we show that it is surprisingly effective for semi-supervised learning on ImageNet. A key ingredient of our approach is the use of big (deep and wide) networks during pretraining and fine-tuning. We find that, the fewer the labels, the more this approach (task-agnostic use of unlabeled data) benefits from a bigger network. After fine-tuning, the big network can be further improved and distilled into a much smaller one with little loss in classification accuracy by using the unlabeled examples for a second time, but in a task-specific way. The proposed semi-supervised learning algorithm can be summarized in three steps: unsupervised pretraining of a big ResNet model using SimCLRv2, supervised fine-tuning on a few labeled examples, and distillation with unlabeled examples for refining and transferring the task-specific knowledge. This procedure achieves 73.9% ImageNet top-1 accuracy with just 1% of the labels (≤13 labeled images per class) using ResNet-50, a 10× improvement in label efficiency over the previous state-of-theart. With 10% of labels, ResNet-50 trained with our method achieves 77.5% top-1 accuracy, outperforming standard supervised training with all of the labels. 1

For agents

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

get_harvested_code_for_paper("2006.10029")
get_code_for_paper("2006.10029")
have("2006.10029")

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

curl https://syntology.ai/api/ran/2006.10029.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.10029.svg)](https://syntology.ai/paper/2006.10029)

Connect an agent — have() is free.