SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.16201 · 2021

MT3: Meta Test-Time Training for Self-Supervised Test-Time Adaption

arXiv · PDF · Open in the Atlas

Code that ran

7 of 9 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 0 have no contract verdict; 5 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/2103.16201.json · badge: /api/ran/2103.16201.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
AlexanderBartler/MT3 canonical (classifier label)
classifier basis: census
7 ran · 2 no recorded run
FunctionStatusWhere it lives
blend Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
AlexanderBartler/MT3/model/augementations/auto_augment.py
code served (permissive licence) · get_code("c6b91a83d159a21c")
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 EagerTensor, ndim 2, dtype family int, shape (3, 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 tensorflow as tf
import numpy as np

# Create small uint8 image tensors
# Image 1: 4x3 (height x width) grayscale image
image1_np = np.array([[50, 100, 150, 200],
                      [60, 110, 160, 210],
                      [70, 120, 170, 220]], dtype=np.uint8)
image1 = tf.constant(image1_np, dtype=tf.uint8)

# Image 2: 4x3 (height x width) grayscale image with different values
image2_np = np.array([[100, 150, 200, 250],
                      [110, 160, 210, 255],
                      [120, 170, 220, 255]], dtype=np.uint8)
image2 = tf.constant(image2_np, dtype=tf.uint8)

# Test with a factor between 0.0 and 1.0 for interpolation
factor = 0.5

ARGS = {
    "image1": image1,
    "image2": image2,
    "factor": factor
}
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 EagerTensor, ndim 2, dtype family int, shape (3, 4) — shape (3, 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
byol_loss_fn Ran
ran, violated the contract we drafted
AlexanderBartler/MT3/model/train_byol.py
code served (permissive licence) · get_code("b1560342ca69b4af")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

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

# Create small sample tensors for x and y
# Shape: (batch_size=2, feature_dim=4)
x = tf.constant(
    np.random.randn(2, 4).astype(np.float32),
    dtype=tf.float32
)

y = tf.constant(
    np.random.randn(2, 4).astype(np.float32),
    dtype=tf.float32
)

ARGS = {
    "x": x,
    "y": y,
}
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 EagerTensor, 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
get_baseline Ran
ran; our contract draft was wrong, not the code
AlexanderBartler/MT3/print_table.py
code served (permissive licence) · get_code("5871a39aaaf17f95")
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 os
import json
import tempfile
import numpy as np

# Create a temporary directory structure with test data
temp_dir = tempfile.mkdtemp()

# Create subdirectories and test_results.json files
for i in range(2):
    subfolder = os.path.join(temp_dir, f'run_{i}')
    os.makedirs(subfolder, exist_ok=True)
    
    test_results = {
        'dataset_a': {'accuracy': '0.85'},
        'dataset_b': {'accuracy': '0.92'},
        'dataset_c': {'accuracy': '0.78'},
    }
    
    results_file = os.path.join(subfolder, 'test_results.json')
    with open(results_file, 'w') as f:
        json.dump(test_results, f)

ARGS = {
    'folder': temp_dir,
    'skip': []
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_data_from_folder Ran
ran; our contract draft was wrong, not the code
AlexanderBartler/MT3/print_table.py
code served (permissive licence) · get_code("d726cce911a1b3b7")
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 os
import json
import tempfile
import numpy as np

# Create a temporary directory structure with JSON files
temp_dir = tempfile.mkdtemp()

# Create subdirectories with result files
subfolder1 = os.path.join(temp_dir, "run1")
subfolder2 = os.path.join(temp_dir, "run2")
os.makedirs(subfolder1)
os.makedirs(subfolder2)

# Create JSON result files
result_data_1 = {
    "dataset_a": {"accuracy": "0.85"},
    "dataset_b": {"accuracy": "0.92"},
    "dataset_c": {"accuracy": "0.78"}
}

result_data_2 = {
    "dataset_a": {"accuracy": "0.87"},
    "dataset_b": {"accuracy": "0.90"},
    "dataset_c": {"accuracy": "0.80"}
}

with open(os.path.join(subfolder1, "results.json"), "w") as f:
    json.dump(result_data_1, f)

with open(os.path.join(subfolder2, "results.json"), "w") as f:
    json.dump(result_data_2, f)

ARGS = {
    "folder": temp_dir,
    "result_name": "results.json",
    "skip": ["dataset_c"]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_short_name Ran
ran; our contract draft was wrong, not the code
AlexanderBartler/MT3/print_table.py
code served (permissive licence) · get_code("6ab763abaaaca859")
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 torch

ARGS = {
    'dataset': 'cifar10',
    'level': 5,
}
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 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
result_meta Ran
ran; our contract draft was wrong, not the code
AlexanderBartler/MT3/model/metrics.py
code served (permissive licence) · get_code("6a5d65d732d4a46c")
How it ran

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

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create mock metric objects that have a .result() method
class MockMetric:
    def __init__(self, value):
        self.value = value
    
    def result(self):
        return self.value

# Create tensors for the metrics
pre_update_accuracy_tensor = torch.tensor([0.85], dtype=torch.float32)
updated_accuracy_tensor = torch.tensor([0.92], dtype=torch.float32)

# Create the metrics dictionary with mock metric objects
metrics_dict = {
    'pre_update_accuracy': MockMetric(pre_update_accuracy_tensor),
    'updated_accuracy': MockMetric(updated_accuracy_tensor)
}

ARGS = {
    'metrics_': metrics_dict,
    'as_numpy': False
}
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 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
solarize Ran
ran, violated the contract we drafted
AlexanderBartler/MT3/model/augementations/auto_augment.py
code served (permissive licence) · get_code("e40dd3a18c0fb8bc")
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 EagerTensor, ndim 3, dtype family int, shape (2, 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 tensorflow as tf
import numpy as np

# Create a small test image as a TensorFlow tensor
# Using uint8 values in range [0, 255] as typical for images
image_data = tf.constant(
    [[[100, 150, 200, 50],
      [255, 128, 75, 25]],
     [[10, 220, 180, 90],
      [64, 32, 200, 255]]], 
    dtype=tf.uint8
)

ARGS = {
    "image": image_data,
    "threshold": 128
}
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 EagerTensor, ndim 3, dtype family int, shape (2, 2, 4) — shape (2, 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
cutout Not yet run
unverified (harvested, no recorded run)
AlexanderBartler/MT3/model/augementations/auto_augment.py
code served (permissive licence) · get_code("e570f106e7008c26")
How it ran

no invocation recorded.

result Not yet run
unverified (harvested, no recorded run)
AlexanderBartler/MT3/model/metrics.py
code served (permissive licence) · get_code("8af693688cedf8ba")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 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
solarize 4 other papers
2 dated earlier, 2 later
1911.04252 (2019-11)
Self-training with Noisy Student improves ImageNet classification
2006.14536 (2020-06)
Smooth Adversarial Training
2110.09192 (2021-10)
Learning Optimal Conformal Classifiers
and 1 more not named here
cutout 3 other papers
2 dated earlier, 1 later
1911.04252 (2019-11)
Self-training with Noisy Student improves ImageNet classification
2006.14536 (2020-06)
Smooth Adversarial Training
2310.13807 (2023-10)
Learning to (Learn at Test Time)
blend 1 other paper
1 later
2110.09192 (2021-10)
Learning Optimal Conformal Classifiers

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

An unresolved problem in Deep Learning is the ability of neural networks to cope with domain shifts during test-time, imposed by commonly fixing network parameters after training. Our proposed method Meta Test-Time Training (MT3), however, breaks this paradigm and enables adaption at test-time. We combine meta-learning, self-supervision and test-time training to learn to adapt to unseen test distributions. By minimizing the self-supervised loss, we learn task-specific model parameters for different tasks. A meta-model is optimized such that its adaption to the different task-specific models leads to higher performance on those tasks. During test-time a single unlabeled image is sufficient to adapt the meta-model parameters. This is achieved by minimizing only the self-supervised loss component resulting in a better prediction for that image. Our approach significantly improves the state-of-the-art results on the CIFAR-10-Corrupted image classification benchmark. Our implementation is available on GitHub.

For agents

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

get_harvested_code_for_paper("2103.16201")
get_code_for_paper("2103.16201")
have("2103.16201")

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

curl https://syntology.ai/api/ran/2103.16201.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.16201.svg)](https://syntology.ai/paper/2103.16201)

Connect an agent — have() is free.