SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2111.07473 · 2021

Scrutinizing XAI using linear ground-truth data with suppressor variables

arXiv · PDF · Open in the Atlas

Code that ran

8 of 8 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 8 have no contract verdict; 0 failures of our instrument; 0 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/2111.07473.json · badge: /api/ran/2111.07473.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
braindatalab/scrutinizing-xai canonical (classifier label)
classifier basis not recorded on the link
8 ran · 0 no recorded run
FunctionStatusWhere it lives
assign_model Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/agnostic_methods/utils.py
code served (permissive licence) · get_code("02590839b3c9ae29")
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 RandomForestClassifier. 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
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression


class DummyModel:
    """A minimal model without predict or predict_proba methods"""
    def __init__(self):
        self.value = 42


class DummySklearnLike:
    """A minimal scikit-learn-like model with predict method"""
    def predict(self, X):
        return np.array([0, 1])


torch.manual_seed(0)

# Create a valid scikit-learn model
sklearn_model = RandomForestClassifier(n_estimators=2, random_state=42)

ARGS = {
    'model': sklearn_model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type RandomForestClassifier — a RandomForestClassifier output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a RandomForestClassifier 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
create_rain_cloud_data Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/visualization.py
code served (permissive licence) · get_code("ee6bd7a4a542ad5c")
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. It ran in syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1) at 2026-09-26T22:17:14Z. 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
import pandas as pd

# Construct the data dictionary structure expected by create_rain_cloud_data
data = {
    'snr_1': {
        'llr': [
            {'roc_auc': [0.85, 0.87, 0.86]},
            {'roc_auc': [0.84, 0.88, 0.85]}
        ],
        'pfi': [
            {'roc_auc': [0.80, 0.82, 0.81]},
            {'roc_auc': [0.79, 0.83, 0.80]}
        ],
        'input_t_gradient': [
            {'roc_auc': [0.75, 0.77, 0.76]}
        ]
    },
    'snr_2': {
        'lime': [
            {'roc_auc': [0.78, 0.80, 0.79]},
            {'roc_auc': [0.77, 0.81, 0.78]}
        ],
        'shap_linear': [
            {'roc_auc': [0.82, 0.84, 0.83]},
            {'roc_auc': [0.81, 0.85, 0.82]}
        ],
        'impurity': [
            {'roc_auc': [0.76, 0.78, 0.77]}
        ]
    },
    'snr_3': {
        'gradient': [
            {'roc_auc': [0.88, 0.90, 0.89]},
            {'roc_auc': [0.87, 0.91, 0.88]}
        ],
        'deep_taylor': [
            {'roc_auc': [0.83, 0.85, 0.84]},
            {'roc_auc': [0.82, 0.86, 0.83]}
        ]
    }
}

ARGS = {
    'data': data,
    'metric_name': 'roc_auc'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, torch — 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 syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1); 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
is_keras_model Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/visualization.py
code served (permissive licence) · get_code("87a611015ca14d59")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Test case 1: 2D array where product equals first dimension (should return False)
data_1 = np.zeros((4, 1))  # shape (4, 1), prod=4, shape[0]=4, so p == shape[0] is True

# Test case 2: 2D array where product equals second dimension (should return False)
data_2 = np.zeros((1, 5))  # shape (1, 5), prod=5, shape[1]=5, so p == shape[1] is True

# Test case 3: 2D array where product equals neither dimension (should return True)
data_3 = np.zeros((3, 4))  # shape (3, 4), prod=12, shape[0]=3, shape[1]=4, neither matches

# Use the third case for the main test
ARGS = {
    "data": data_3
}
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 bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
is_sample_based Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/visualization.py
code served (permissive licence) · get_code("738bd82d36dd6a1c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# The function checks if a numpy array has a specific shape property
# It returns True if the product of shape dimensions is NOT equal to any single dimension
# This helps identify multi-dimensional sample-based data

# Case 1: A 2D array where product of shape equals one of the dimensions
# shape = (4, 3), product = 12, not equal to 4 or 3 -> returns True
data_2d = np.random.randn(4, 3)

ARGS = {
    "data": data_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 bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
is_scikit_learn_model Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/agnostic_methods/utils.py
code served (permissive licence) · get_code("b672e90f826d3720")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from typing import Any

# Create a mock scikit-learn-like model with predict method
class MockSklearnModel:
    def predict(self, X):
        return [0, 1]

# Create a mock scikit-learn-like model with predict_proba method
class MockSklearnModelProba:
    def predict_proba(self, X):
        return [[0.1, 0.9], [0.8, 0.2]]

# Create a model without scikit-learn methods
class NonSklearnModel:
    def forward(self, X):
        return X

# Test with different model types
model_with_predict = MockSklearnModel()
model_with_predict_proba = MockSklearnModelProba()
non_sklearn_model = NonSklearnModel()

ARGS = {
    "model": model_with_predict
}
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 bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_json_file Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/common.py
code served (permissive licence) · get_code("bcc07735b7a488c3")
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 tempfile
import os

# Create a temporary JSON file with valid content
temp_dir = tempfile.mkdtemp()
temp_file_path = os.path.join(temp_dir, "test.json")

test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "dict"}
}

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

ARGS = {
    "file_path": temp_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
load_pickle Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/common.py
code served (permissive licence) · get_code("7fe7bf864f67a7e8")
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 pickle
import tempfile
import os

# Create a temporary pickle file with some test data
test_data = {
    'key1': 'value1',
    'key2': [1, 2, 3],
    'key3': {'nested': 'dict'}
}

# Create a temporary file
temp_file = tempfile.NamedTemporaryFile(mode='wb', delete=False, suffix='.pkl')
pickle.dump(test_data, temp_file)
temp_file.close()

# Define ARGS with the file path
ARGS = {
    'file_path': temp_file.name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
to_pickle Ran
ran on a synthesized input
braindatalab/scrutinizing-xai/common.py
code served (permissive licence) · get_code("97524ba0a8c15bf6")
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 os
import pickle
import tempfile
from os.path import join

ARGS = {
    'output_dir': tempfile.mkdtemp(),
    'data': {
        'key1': 'value1',
        'key2': [1, 2, 3],
        'key3': {'nested': 'dict'}
    },
    'suffix': 'test_output'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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

Machine learning (ML) is increasingly often used to inform high-stakes decisions. As complex ML models (e.g., deep neural networks) are often considered black boxes, a wealth of procedures has been developed to shed light on their inner workings and the ways in which their predictions come about, defining the field of 'explainable AI' (XAI). Saliency methods rank input features according to some measure of 'importance'. Such methods are difficult to validate since a formal definition of feature importance is, thus far, lacking. It has been demonstrated that some saliency methods can highlight features that have no statistical association with the prediction target (suppressor variables). To avoid misinterpretations due to such behavior, we propose the actual presence of such an association as a necessary condition and objective preliminary definition for feature importance. We carefully crafted a ground-truth dataset in which all statistical dependencies are well-defined and linear, serving as a benchmark to study the problem of suppressor variables. We evaluate common explanation methods including LRP, DTD, PatternNet, PatternAttribution, LIME, Anchors, SHAP, and permutation-based methods with respect to our objective definition. We show that most of these methods are unable to distinguish important features from suppressors in this setting.

For agents

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

get_harvested_code_for_paper("2111.07473")
get_code_for_paper("2111.07473")
have("2111.07473")

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

curl https://syntology.ai/api/ran/2111.07473.json

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

[![Syntology run record](https://syntology.ai/api/ran/2111.07473.svg)](https://syntology.ai/paper/2111.07473)

Connect an agent — have() is free.