SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.01351 · CVPR · 2024

AETTA: Label-Free Accuracy Estimation for Test-Time Adaptation

Taesik Gong, Sung-Ju Lee, Taeckyung Lee, Sorn Chottananurak

arXiv · PDF · Open in the Atlas

Code that ran

12 of 12 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 9 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/2404.01351.json · badge: /api/ran/2404.01351.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
taeckyung/AETTA canonical (classifier label)
classifier basis: census
12 ran · 0 no recorded run
FunctionStatusWhere it lives
calc_energy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
taeckyung/AETTA/utils/loss_functions.py
code served (permissive licence) · get_code("fc962c3df9bd26ac")
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 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

torch.manual_seed(0)

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

# Set temperature factor
temp_factor = 1.0

# Construct the arguments dictionary
ARGS = {
    "x": x,
    "temp_factor": temp_factor,
}
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 1, dtype family float — no output shape is stated: observe_signature.py could not observe it (import:OSError)
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_avg_online_acc Ran
ran, honoured the contract we drafted
taeckyung/AETTA/print_acc.py
code served (permissive licence) · get_code("748df4ce43772fae")
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 float, 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 json
import tempfile
import os

# Create a temporary JSON file with the expected structure
temp_dir = tempfile.mkdtemp()
json_file_path = os.path.join(temp_dir, "test_data.json")

# Write test data to the JSON file
test_data = {
    "accuracy": [0.5, 0.6, 0.7, 0.8, 0.85]
}

with open(json_file_path, 'w') as f:
    json.dump(test_data, f)

ARGS = {
    "file_path": json_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 float, 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
get_avg_online_acc Ran
ran, honoured the contract we drafted
taeckyung/AETTA/print_est.py
code served (permissive licence) · get_code("0764e24b7e85a924")
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 json
import tempfile
import numpy as np

# Create a temporary JSON file with the required data structure
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False)
json_data = {
    'current_accuracy': [0.75, 0.82, 0.88, 0.91],
    'aetta': [0.76, 0.81, 0.87, 0.90],
    'other_metric': [75.5, 82.3, 87.1, 91.2]
}
json.dump(json_data, temp_file)
temp_file.close()

ARGS = {
    'file_path': temp_file.name,
    'target': 'aetta'
}
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_dirnames Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/process_cifar.py
code served (permissive licence) · get_code("7d3f76502e1e286c")
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

ARGS = {
    'dataset': 'cifar-10c'
}
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, 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_normalize_layer Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/utils/normalize_layer.py
code served (permissive licence) · get_code("2b4bb33f8c8c082c")
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 NormalizeLayer. 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

torch.manual_seed(0)

ARGS = {
    'dataset': 'cifar10'
}
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 NormalizeLayer — a NormalizeLayer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NormalizeLayer 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
get_normalize_std Ran
ran, honoured the contract we drafted
taeckyung/AETTA/utils/normalize_layer.py
code served (permissive licence) · get_code("aeeb23c68bb4b797")
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 torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    'dataset': 'cifar10'
}
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 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_data Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/process_cifar.py
pointer only (licence: MIT); the stored copy get_code serves, from dmn-sjk/ar-tta, is pointer only (licence: NONE) · get_code("8e952e7f9820216a")
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 dmn-sjk/ar-tta, 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.

marginal_entropy Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/utils/loss_functions.py
code served (permissive licence) · get_code("f194bfd61755bba9")
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 numpy as np

torch.manual_seed(0)

# The function takes outputs as input
# outputs is expected to be a tensor that:
# 1. Supports logsumexp operation with dim=-1 and keepdim=True
# 2. Has shape[0] accessible (batch dimension)
# 3. Supports logsumexp with dim=0
# 4. Supports exp() operation
# 5. Supports sum(dim=-1) operation

# Create a small batch of logits
# Shape: (batch_size=2, num_classes=4)
outputs = torch.randn(2, 4)

ARGS = {
    "outputs": outputs
}
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 — 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
process_path Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/print_acc.py
code served (permissive licence) · get_code("7483be869cdb6276")
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 json
import os
import tempfile
import shutil
from pathlib import Path

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

# Create mock directory structure and files
def create_mock_structure(base_path):
    """Create a mock directory structure with online_eval.json files"""
    corruptions = ['gaussian_noise', 'shot_noise', 'impulse_noise']
    seeds = [0, 1]
    outdists = ['id', 'ood']
    
    for seed in seeds:
        for outdist in outdists:
            for corr in corruptions:
                # Create directory path
                dir_path = os.path.join(
                    base_path,
                    f'{corr}_{outdist}',
                    'reset_1',
                    f'prefix_seed{seed}_dist0.1'
                )
                os.makedirs(dir_path, exist_ok=True)
                
                # Create mock online_eval.json
                mock_data = {'accuracy': [0.5, 0.6, 0.75]}
                with open(os.path.join(dir_path, 'online_eval.json'), 'w') as f:
                    json.dump(mock_data, f)

create_mock_structure(temp_dir)

# Create a simple args class
class Args:
    def __init__(self):
        self.seed = [0, 1]
        self.outdist = ['id', 'ood']
        self.cont = False
        self.suffix = 'suffix'
        self.reset = '1'
        self.prefix = 'prefix'
        self.dataset = 'cifar10outdist'
        self.dist = 0.1

args = Args()
path = temp_dir

ARGS = {
    'args': args,
    'path': path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — 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
process_path Ran
ran; our contract draft was wrong, not the code
taeckyung/AETTA/print_est.py
code served (permissive licence) · get_code("f170a751d5d3e5aa")
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 json
import os
import tempfile
import shutil
from pathlib import Path

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

# Create nested directory structure with mock online_eval.json files
def create_mock_structure():
    """Create a mock directory structure with test data"""
    base_path = os.path.join(temp_dir, 'test_method_outdist')
    os.makedirs(base_path, exist_ok=True)
    
    # Create subdirectories matching expected patterns
    seeds = [0, 1]
    outdists = ['in', 'out']
    corruptions = ['gaussian_noise', 'shot_noise', 'defocus_blur']
    
    for seed in seeds:
        for outdist in outdists:
            for corr in corruptions:
                if outdist == 'in':
                    subdir = os.path.join(base_path, f'{corr}_{outdist}', f'test_prefix_{seed}_dist1_{corr}_data')
                else:
                    subdir = os.path.join(base_path, f'{corr}_{outdist}', f'test_prefix_{seed}_dist1_{corr}_data')
                os.makedirs(subdir, exist_ok=True)
                
                # Create mock online_eval.json
                mock_data = {
                    'current_accuracy': [0.7, 0.75, 0.8, 0.82],
                    'aetta': [0.71, 0.76, 0.81, 0.83]
                }
                json_path = os.path.join(subdir, 'online_eval.json')
                with open(json_path, 'w') as f:
                    json.dump(mock_data, f)
    
    return base_path

test_path = create_mock_structure()

# Create mock args object
class Args:
    def __init__(self):
        self.seed = [0, 1]
        self.outdist = ['in', 'out']
        self.cont = False
        self.reset = None
        self.dataset = 'cifar10'
        self.prefix = 'test_prefix'
        self.dist = 1
        self.suffix = ''
        self.target = 'aetta'

ARGS = {
    'args': Args(),
    'path': test_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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
reshape Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
taeckyung/AETTA/process_cifar.py
pointer only (licence: MIT); the stored copy get_code serves, from dmn-sjk/ar-tta, is pointer only (licence: NONE) · get_code("92bc8db96a48bcb0")
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 dmn-sjk/ar-tta, 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.

softmax_entropy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
taeckyung/AETTA/utils/loss_functions.py
code served (permissive licence) · get_code("a9edd84f4eedeec8")
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 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

torch.manual_seed(0)

# Create a small batch of logits
# Shape: (batch_size=2, num_classes=4)
x = torch.randn(2, 4)

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 1, dtype family float — no output shape is stated: observe_signature.py could not observe it (import:OSError)
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

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 7 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 4 later) and 5 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
marginal_entropy 3 other papers
1 dated earlier, 2 later
2310.18816 (2023-10)
Adaptive Test-Time Personalization for Federated Learning
2407.12128 (2024-07)
Distribution Alignment for Fully Test-Time Adaptation with Dynamic Online Data Streams
2609.02507 (2026-09)
Rethinking the Teacher-Student Framework for Test-Time Adaptation
load_data 2 other papers
1 dated earlier, 1 later
2309.10109 (2023-09)
AR-TTA: A Simple Method for Real-World Continual Test-Time Adaptation
2505.18514 (2025-05)
Test-Time Adaptation with Binary Feedback
reshape 2 other papers
1 dated earlier, 1 later
2309.10109 (2023-09)
AR-TTA: A Simple Method for Real-World Continual Test-Time Adaptation
2505.18514 (2025-05)
Test-Time Adaptation with Binary Feedback
calc_energy 1 other paper
1 dated earlier
2310.10074 (2023-10)
SoTTA: Robust Test-Time Adaptation on Noisy Data Streams
get_avg_online_acc 1 other paper
1 dated earlier
2208.05117 (2022-08)
NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation
get_dirnames 1 other paper
1 later
2505.18514 (2025-05)
Test-Time Adaptation with Binary Feedback
softmax_entropy 1 other paper
1 dated earlier
2310.10074 (2023-10)
SoTTA: Robust Test-Time Adaptation on Noisy Data Streams

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Test-time adaptation (TTA) has emerged as a viable solution to adapt pre-trained models to domain shifts using unlabeled test data. However, TTA faces challenges of adaptation failures due to its reliance on blind adaptation to unknown test samples in dynamic scenarios. Traditional methods for out-of-distribution performance estimation are limited by unrealistic assumptions in the TTA context, such as requiring labeled data or re-training models. To address this issue, we propose AETTA, a label-free accuracy estimation algorithm for TTA. We propose the prediction disagreement as the accuracy estimate, calculated by comparing the target model prediction with dropout inferences. We then improve the prediction disagreement to extend the applicability of AETTA under adaptation failures. Our extensive evaluation with four baselines and six TTA methods demonstrates that AETTA shows an average of 19.8%p more accurate estimation compared with the baselines. We further demonstrate the effectiveness of accuracy estimation with a model recovery case study, showcasing the practicality of our model recovery based on accuracy estimation. The source code is available at https://github.com/taeckyung/AETTA.

For agents

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

get_harvested_code_for_paper("2404.01351")
get_code_for_paper("2404.01351")
have("2404.01351")

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

curl https://syntology.ai/api/ran/2404.01351.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.01351.svg)](https://syntology.ai/paper/2404.01351)

Connect an agent — have() is free.