SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.08958 · 2022

On the Usefulness of the Fit-on-the-Test View on Evaluating Calibration of Classifiers

arXiv · PDF · Open in the Atlas

Code that ran

17 of 22 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, 1 violated it and 15 have no contract verdict; 1 failure of our instrument; 5 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/2203.08958.json · badge: /api/ran/2203.08958.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
markus93/fit-on-the-test canonical (classifier label)
classifier basis not recorded on the link
17 ran · 5 no recorded run
FunctionStatusWhere it lives
construct_calibration_fun Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/calibration_function_derivates.py
code served (permissive licence) · get_code("8f150358dd1a028b")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# The function construct_calibration_fun takes:
# 1. calibration_fun: a callable that takes x and returns a transformed version
# 2. fun_amount: a float between 0 and 1 that controls the mix

# Create a simple calibration function that applies a transformation to input
def calibration_fun(x):
    # Simple calibration: scale and shift
    return x * 2.0 + 0.5

# fun_amount is a float controlling the interpolation between identity and calibration
fun_amount = 0.3

ARGS = {
    "calibration_fun": calibration_fun,
    "fun_amount": fun_amount,
}
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 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
construct_correlation_data_row Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/dataframe_helpers.py
code served (permissive licence) · get_code("f669081dc30a1915")
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 pandas as pd
import numpy as np
from scipy import stats

# Create a small pandas DataFrame with the required columns
n_samples = 10
selection = pd.DataFrame({
    'c_hat_distance_p': np.random.randn(n_samples),
    'p_distance_c': np.random.randn(n_samples),
    'n_data': [100] * n_samples,
    'binning': ['uniform'] * n_samples,
    'n_bins': [5] * n_samples,
    'error_type': ['mse'] * n_samples,
})

ARGS = {
    'selection': selection
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, scipy — 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
construct_data_row Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Synthetic/run_method.py
code served (permissive licence) · get_code("0b75185643864e8d")
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 numpy as np

torch.manual_seed(0)

# Construct arguments for construct_data_row
ARGS = {
    'n_folds': 5,
    'binning_name': 'uniform',
    'n_bins': 10,
    'c_hat_distance_p': 0.123,
    'c_hat_distance_p_debiased': 0.118,
    'c_hat_distance_p_square': 0.0151,
    'c_hat_distance_p_square_debiased': 0.0139,
    'c_hat_distance_c': 0.145,
    'c_hat_distance_c_square': 0.0210,
    'test_c_hat_distance_p': 0.135,
    'test_c_hat_distance_p_debiased': 0.129,
    'test_c_hat_distance_p_square': 0.0182,
    'test_c_hat_distance_p_square_debiased': 0.0166,
    'test_c_hat_distance_c': 0.157,
    'test_c_hat_distance_c_square': 0.0247,
}
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 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
construct_data_row_from_binning Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/dataframe_helpers.py
code served (permissive licence) · get_code("051fb62188a6ca48")
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 numpy as np

class MockBinning:
    """Stand-in class for binning object with required attributes."""
    def __init__(self):
        self.binning_name = "equal_width"
        self.n_bins = 10
        self.c_hat_distance_p_square = 0.0523
        self.c_hat_distance_p_abs = 0.0412

ARGS = {
    'binning': MockBinning(),
    'calibration_function_name': 'temperature_scaling',
    'true_calibration_error': 0.1234,
    'seed': 42,
    'n_data': 1000,
    'error_type': 'square',
    'expected_calibration_error': 0.0987
}
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 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
find_expected_calibration_error Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/calibration_function_derivates.py
code served (permissive licence) · get_code("55ae28f90999512f")
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 float64, ndim 0, 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 numpy as np
from scipy.stats import beta

# Define a simple function that maps [0,1] to [0,1]
# This represents a model's confidence calibration
def calibration_function(x):
    """A simple sigmoid-like function for calibration"""
    return 1 / (1 + np.exp(-5 * (x - 0.5)))

# Define an error function that computes squared error
def error_func(residual):
    """Squared error of the residual"""
    return residual ** 2

# Construct ARGS dictionary
ARGS = {
    "function": calibration_function,
    "error_function": error_func,
    "beta_alpha": 1,
    "beta_beta": 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
find_suitable_fun_amount Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/calibration_function_derivates.py
code served (permissive licence) · get_code("79d72a8e7def9af5")
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 float64, ndim 0, 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 numpy as np
from scipy.stats import beta

# Define a simple calibration function (e.g., sigmoid-like behavior)
def calibration_fun(x):
    """A simple calibration function that maps [0,1] to [0,1]"""
    return np.clip(0.5 + 0.3 * np.sin(np.pi * x), 0, 1)

# Define an error function (e.g., absolute error)
def error_fun(x):
    """Error function - computes absolute error"""
    return np.abs(x)

# Define the target calibration error
true_calibration_error = 0.05

# Beta distribution parameters
beta_alpha = 1
beta_beta = 1

ARGS = {
    "calibration_fun": calibration_fun,
    "true_calibration_error": true_calibration_error,
    "error_fun": error_fun,
    "beta_alpha": beta_alpha,
    "beta_beta": beta_beta,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
generate_data Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/data_generation.py
code served (permissive licence) · get_code("0e19229d023ae722")
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 2, 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 numpy as np

def calibration_function(pred):
    """Simple calibration function that applies a transformation to predictions."""
    return pred * 0.9 + 0.05

ARGS = {
    "dirichlet": [1.0, 1.0, 1.0],
    "n_data": 10,
    "calibration_function": calibration_function,
    "seed": 42
}
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 2, 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_cgts Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/df_unify_5m.py
code served (permissive licence) · get_code("71335b4eb90f29e0")
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 DataFrame. 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 numpy as np
import pandas as pd

# Create sample data for the DataFrame
# The function expects 4 columns with list/array-like values
n_rows = 2
n_elements = 4

# Create lists of arrays for each column
c_hat_distance_c = [np.random.randn(n_elements) for _ in range(n_rows)]
c_hat_distance_c_square = [np.random.randn(n_elements) for _ in range(n_rows)]
p_distance_c = [np.random.randn(n_elements) for _ in range(n_rows)]
p_distance_c_square = [np.random.randn(n_elements) for _ in range(n_rows)]

# Create DataFrame with the required columns
df = pd.DataFrame({
    'c_hat_distance_c': c_hat_distance_c,
    'c_hat_distance_c_square': c_hat_distance_c_square,
    'p_distance_c': p_distance_c,
    'p_distance_c_square': p_distance_c_square,
    'other_column': [1, 2]  # Additional column to be preserved
})

ARGS = {
    'df': df
}
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 DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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_data_name Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/df_unify_5m.py
code served (permissive licence) · get_code("2d6d5280ab3aec05")
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
import torch.nn as nn

torch.manual_seed(0)

# The function checks if certain strings are in the 'file' parameter
# It performs string containment checks only, so we just need a string

ARGS = {
    'file': 'resnet110_checkpoint.pth'
}
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
log_encode Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/cal_methods.py
code served (permissive licence) · get_code("e742a88077f30c11")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (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 numpy as np

# Create a small numpy array as input
# The function expects a numpy array-like object with dtype and clipping/logging operations
x = np.array([0.1, 0.5, 0.9, 0.3], dtype=np.float32)

ARGS = {
    "x": x
}
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 ndarray, ndim 1, dtype family float, shape (4,) — shape (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
mirror_1d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
markus93/fit-on-the-test/Experiments_Pseudo/kde.py
code served (permissive licence) · get_code("8f3819e94710b239")
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 ndarray, ndim 2, dtype family float, shape (8, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small 1D array for testing
d = np.array([[1.0], [2.0], [3.0], [4.0]])

# Define arguments for mirror_1d
ARGS = {
    "d": d,
    "xmin": 0.5,
    "xmax": 4.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (8, 1) — shape (8, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
poly Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/calibration_functions.py
code served (permissive licence) · get_code("7f8c2b881b22648e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (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 numpy as np
import torch

# Analyze the function: poly(x) takes a single argument x and passes it to
# np.polynomial.polynomial.polyval(x, [0, 4.09, -17.62, 35.17, -32.2, 11.56])
# polyval evaluates a polynomial at x, where x can be a scalar, array, or array-like

# Create a small numpy array as input
x = np.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    'x': x
}
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 ndarray, ndim 1, dtype family float, shape (4,) — shape (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
select Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/dataframe_helpers.py
code served (permissive licence) · get_code("115a8789d9377db2")
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 DataFrame. 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 pandas as pd
import numpy as np

# Create a sample DataFrame with the required columns
df = pd.DataFrame({
    'n_data': [100, 200, 100, 300, 200],
    'binning': ['uniform', 'log', 'uniform', 'log', 'uniform'],
    'error_type': ['gaussian', 'poisson', 'gaussian', 'poisson', 'gaussian'],
    'value': [1.5, 2.3, 3.1, 4.2, 5.0]
})

ARGS = {
    'df': df,
    'n_data': 100,
    'binning': 'uniform',
    'error_type': 'gaussian'
}
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 DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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 Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/cal_methods.py
code served (permissive licence) · get_code("80483261c3f6087d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 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 numpy as np

# Create a small 2D numpy array (2 samples, 4 dimensions)
x_2d = np.array([
    [1.0, 2.0, 3.0, 4.0],
    [2.0, 1.0, 0.5, 1.5]
], dtype=np.float32)

ARGS = {
    "x": x_2d
}
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 ndarray, 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
split_data Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/main_NN_1m_final.py
code served (permissive licence) · get_code("4dc7cd36d2936a07")
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. 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 numpy as np

# Create small sample data arrays with distinct sizes
# The function expects a list of arrays where the last one determines the data length
input_array_1 = np.random.rand(100, 4)  # 100 samples, 4 features
input_array_2 = np.random.rand(100, 3)  # 100 samples, 3 features
input_array_3 = np.random.rand(100, 5)  # 100 samples, 5 features (last input)

inputs = [input_array_1, input_array_2, input_array_3]

ARGS = {
    "inputs": inputs,
    "split_size": 1000,
    "seed": 5
}
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 list — a list 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
sqrt Ran
ran on a synthesized input
markus93/fit-on-the-test/Experiments_Pseudo/calibration_functions.py
code served (permissive licence) · get_code("645602835255b5f9")
How it ran

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

import torch
import numpy as np

torch.manual_seed(0)

# The sqrt function takes a single parameter x and returns x ** 0.5
# x can be a number, numpy array, or torch tensor
# We'll create a small tensor to demonstrate

x = torch.tensor([[1.0, 4.0, 9.0, 16.0],
                  [25.0, 36.0, 49.0, 64.0]])

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 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
square Ran
ran, violated the contract we drafted
markus93/fit-on-the-test/Experiments_Pseudo/calibration_functions.py
pointer only (licence: MIT); the stored copy get_code serves, from jinglingli/nn-extrapolate, is pointer only (licence: NONE) · get_code("83be324d300a4045")
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 jinglingli/nn-extrapolate, 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.

binning_n_bins_with_crossvalidation Not yet run
unverified (harvested, no recorded run)
markus93/fit-on-the-test/Experiments_Pseudo/binnings.py
code served (permissive licence) · get_code("4139ecf4c8942eb3")
How it ran

no invocation recorded.

construct_data_row_from_binning Not yet run
unverified (harvested, no recorded run)
markus93/fit-on-the-test/Experiments_Synthetic/run_method.py
code served (permissive licence) · get_code("bb7567171178b1d4")
How it ran

no invocation recorded.

construct_data_row_from_raw_data Not yet run
unverified (harvested, no recorded run)
markus93/fit-on-the-test/Experiments_Synthetic/run_method.py
code served (permissive licence) · get_code("debb147433f25086")
How it ran

no invocation recorded.

get_preds_all Not yet run
unverified (harvested, no recorded run)
markus93/fit-on-the-test/Experiments_Pseudo/cal_methods.py
code served (permissive licence) · get_code("8302116db328969a")
How it ran

no invocation recorded.

get_strs Not yet run
unverified (harvested, no recorded run)
markus93/fit-on-the-test/Experiments_Pseudo/df_unify_5m.py
code served (permissive licence) · get_code("095a6e6431202e35")
How it ran

no invocation recorded.

Also present in other papers' code

Of 22 samples lifted from this paper's linked repositories, 22 carry a function identity and 0 do not. Of the 22 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 20 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
mirror_1d 1 other paper
1 dated earlier
2003.07329 (2020-03)
Mix-n-Match: Ensemble and Compositional Methods for Uncertainty Calibration in Deep Learning
square 1 other paper
1 dated earlier
2009.11848 (2020-09)
How Neural Networks Extrapolate: From Feedforward to Graph Neural Networks

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

Every uncalibrated classifier has a corresponding true calibration map that calibrates its confidence. Deviations of this idealistic map from the identity map reveal miscalibration. Such calibration errors can be reduced with many post-hoc calibration methods which fit some family of calibration maps on a validation dataset. In contrast, evaluation of calibration with the expected calibration error (ECE) on the test set does not explicitly involve fitting. However, as we demonstrate, ECE can still be viewed as if fitting a family of functions on the test data. This motivates the fit-on-the-test view on evaluation: first, approximate a calibration map on the test data, and second, quantify its distance from the identity. Exploiting this view allows us to unlock missed opportunities: (1) use the plethora of post-hoc calibration methods for evaluating calibration; (2) tune the number of bins in ECE with cross-validation. Furthermore, we introduce: (3) benchmarking on pseudo-real data where the true calibration map can be estimated very precisely; and (4) novel calibration and evaluation methods using new calibration map families PL and PL3.

For agents

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

get_harvested_code_for_paper("2203.08958")
get_code_for_paper("2203.08958")
have("2203.08958")

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

curl https://syntology.ai/api/ran/2203.08958.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.08958.svg)](https://syntology.ai/paper/2203.08958)

Connect an agent — have() is free.