SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.04518 · 2026

Reproducibility study on how to find Spurious Correlations, Shortcut Learning, Clever Hans or Group-Distributional nonrobustness and how to fix them

Sidney Bender, Ole Delzer

arXiv · PDF

Code that ran

9 of 23 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 0 have no contract verdict; 9 failures of our instrument; 14 no recorded run), measured 2026-11-22; 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/2604.04518.json · badge: /api/ran/2604.04518.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
kohpangwei/group_DRO canonical (classifier label)
classifier basis: census
6 ran · 4 no recorded run
frederikpahde/rrclarc canonical (classifier label)
classifier basis: census
3 ran · 10 no recorded run
FunctionStatusWhere it lives
acc_and_f1 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
kohpangwei/group_DRO/utils_glue.py
pointer only (licence: MIT); the stored copy get_code serves, from danrsc/bert_brain_neurips_2019, is pointer only (licence: NONE) · get_code("cac113ca87b9d9f3")
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 danrsc/bert_brain_neurips_2019, 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.

accuracy Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
kohpangwei/group_DRO/utils.py
code served (permissive licence) · get_code("621de5dd9c62c06d")
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 2, dtype family float; no output shape is stated. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn

torch.manual_seed(0)

# Create small batch of output logits (batch_size=2, num_classes=4)
output = torch.randn(2, 4)

# Create small batch of target labels (batch_size=2)
target = torch.tensor([1, 3])

# Define topk parameter
topk = (1,)

ARGS = {
    "output": output,
    "target": target,
    "topk": topk
}
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 list, ndim 2, 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
convert_examples_to_features Ran
ran; our contract draft was wrong, not the code
kohpangwei/group_DRO/utils_glue.py
code served (permissive licence) · get_code("c4c16e96906f44c1")
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 logging

logging.basicConfig(level=logging.INFO)

class SimpleTokenizer:
    """Mock tokenizer for testing."""
    def __init__(self):
        self.vocab = {
            '[CLS]': 101,
            '[SEP]': 102,
            '[PAD]': 0,
            'hello': 1,
            'world': 2,
            'this': 3,
            'is': 4,
            'a': 5,
            'test': 6,
            'example': 7,
            'text': 8,
            'good': 9,
            'bad': 10,
        }
    
    def tokenize(self, text):
        """Tokenize text into tokens."""
        return text.lower().split()
    
    def convert_tokens_to_ids(self, tokens):
        """Convert tokens to IDs."""
        return [self.vocab.get(token, 1) for token in tokens]


class Example:
    """Simple example class."""
    def __init__(self, guid, text_a, text_b, label):
        self.guid = guid
        self.text_a = text_a
        self.text_b = text_b
        self.label = label


# Create simple examples
examples = [
    Example('1', 'hello world', 'this is test', 'good'),
    Example('2', 'example text', None, 'bad'),
]

# Create label list
label_list = ['good', 'bad']

# Create tokenizer
tokenizer = SimpleTokenizer()

# Set up ARGS
ARGS = {
    'examples': examples,
    'label_list': label_list,
    'max_seq_length': 32,
    'tokenizer': tokenizer,
    'output_mode': 'classification',
    'cls_token_at_end': False,
    'cls_token': '[CLS]',
    'cls_token_segment_id': 0,
    'sep_token': '[SEP]',
    'sep_token_extra': False,
    'pad_on_left': False,
    'pad_token': 0,
    'pad_token_segment_id': 0,
    'sequence_a_segment_id': 0,
    'sequence_b_segment_id': 1,
    'mask_padding_with_zero': True,
}
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 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
extract_wnid Ran
ran; our contract draft was wrong, not the code
frederikpahde/rrclarc/datasets/imagenet.py
code served (permissive licence) · get_code("8ddc981d69ea3f87")
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 numpy as np

torch.manual_seed(0)

def extract_wnid(path):
    return path.split('/')[-1]

# The function takes a single parameter 'path' which is a string
# It splits by '/' and returns the last element
# So we need to provide a path string with '/' separators

ARGS = {
    'path': 'some/directory/structure/wnid12345'
}
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 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
get_accs_for_epoch_across_batches Ran
ran; our contract draft was wrong, not the code
kohpangwei/group_DRO/analysis_utils.py
code served (permissive licence) · get_code("381380e09fd105f8")
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 numpy as np
import pandas as pd

# Create a small DataFrame that matches the expected structure
# The function expects:
# - 'epoch' column
# - 'avg_acc_group:{group}' columns for each group
# - 'processed_data_count_group:{group}' columns for each group

n_groups = 3
n_rows = 4

data = {
    'epoch': [1, 1, 1, 2],
}

# Add accuracy and count columns for each group
for group in range(n_groups):
    data[f'avg_acc_group:{group}'] = np.random.rand(n_rows)
    data[f'processed_data_count_group:{group}'] = np.random.randint(10, 50, n_rows)

df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'epoch': 1,
}
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 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_loss Ran
ran; our contract draft was wrong, not the code
frederikpahde/rrclarc/model_training/training_utils.py
code served (permissive licence) · get_code("ced481b63e75a562")
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 CrossEntropyLoss. 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)

# The function takes loss_name (str) and optional weights (tensor for CrossEntropyLoss)
# CrossEntropyLoss accepts weight as a 1D tensor of per-class weights
# We'll create a small weights tensor for a 4-class classification problem

weights = torch.tensor([1.0, 2.0, 1.5, 0.5])

ARGS = {
    'loss_name': 'cross_entropy',
    'weights': weights
}
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 CrossEntropyLoss — a CrossEntropyLoss output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a CrossEntropyLoss 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_optimizer Ran
ran; our contract draft was wrong, not the code
frederikpahde/rrclarc/model_training/training_utils.py
code served (permissive licence) · get_code("e0a13eb48a172c2d")
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 Adam. 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 a small model with parameters
model = nn.Linear(4, 3)

# Extract model parameters (this is what the optimizer needs)
params = model.parameters()

# Define arguments for get_optimizer
ARGS = {
    'optim_name': 'adam',
    'params': params,
    'lr': 0.001
}
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 Adam — a Adam output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Adam 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
sanitize_df Ran
ran; our contract draft was wrong, not the code
kohpangwei/group_DRO/analysis_utils.py
code served (permissive licence) · get_code("8a66892616008c7f")
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 valid DataFrame with epoch and batch columns
# The function expects a df with at least 'epoch' and 'batch' columns
data = {
    'epoch': [0, 0, 0, 1, 1, 1, 2, 2],
    'batch': [0, 1, 2, 0, 1, 2, 0, 1],
    'loss': [0.5, 0.4, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05]
}
df = pd.DataFrame(data)

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
tokenize Ran
ran; our contract draft was wrong, not the code
kohpangwei/group_DRO/dataset_scripts/generate_multinli.py
code served (permissive licence) · get_code("7ef9ae2541d2ae3e")
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 string

def tokenize(s):
    s = s.translate(str.maketrans('', '', string.punctuation))
    s = s.lower()
    s = s.split(' ')
    return s

ARGS = {
    's': "Hello, World! This is a test."
}
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 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
combine_and_mask Not yet run
unverified (harvested, no recorded run)
kohpangwei/group_DRO/dataset_scripts/dataset_utils.py
code served (permissive licence) · get_code("fe49a67591e2b166")
How it ran

no invocation recorded.

crop_and_resize Not yet run
unverified (harvested, no recorded run)
kohpangwei/group_DRO/dataset_scripts/dataset_utils.py
code served (permissive licence) · get_code("e2f8b1a98dec9723")
How it ran

no invocation recorded.

get_efficientnet Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/efficientnet.py
code served (permissive licence) · get_code("a34ee1b3b27a6aab")
How it ran

no invocation recorded.

get_efficientnet_b0 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/efficientnet.py
code served (permissive licence) · get_code("cb34892eb3e6a222")
How it ran

no invocation recorded.

get_efficientnet_b4 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/efficientnet.py
code served (permissive licence) · get_code("0d137e4287625df2")
How it ran

no invocation recorded.

get_resnet18 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/resnet.py
code served (permissive licence) · get_code("2c22c149a0985110")
How it ran

no invocation recorded.

get_resnet34 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/resnet.py
code served (permissive licence) · get_code("06ad82cc77eece0b")
How it ran

no invocation recorded.

get_resnet50 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/resnet.py
code served (permissive licence) · get_code("997028fdd788415d")
How it ran

no invocation recorded.

get_vgg13 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/vgg.py
code served (permissive licence) · get_code("740913d7e078246b")
How it ran

no invocation recorded.

get_vgg16 Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/vgg.py
code served (permissive licence) · get_code("08582c38903e9dc0")
How it ran

no invocation recorded.

get_vgg16_bn Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/models/vgg.py
code served (permissive licence) · get_code("0d63c1c1b235a76e")
How it ran

no invocation recorded.

load_log Not yet run
unverified (harvested, no recorded run)
kohpangwei/group_DRO/analysis_utils.py
code served (permissive licence) · get_code("10e220982fed5f01")
How it ran

no invocation recorded.

run_one_epoch Not yet run
unverified (harvested, no recorded run)
frederikpahde/rrclarc/model_training/train_model.py
code served (permissive licence) · get_code("26e3775d1e8a65d0")
How it ran

no invocation recorded.

simple_accuracy Not yet run
unverified (harvested, no recorded run)
kohpangwei/group_DRO/utils_glue.py
code served (permissive licence) · get_code("3c241ecfe3749a6d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 23 samples lifted from this paper's linked repositories, 23 carry a function identity and 0 do not. Of the 23 distinct functions with an identity, 12 are also present in other papers whose code we hold (12 with at least one paper dated earlier, 0 the same month, 0 later) and 11 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
simple_accuracy 59 other papers
59 dated earlier
1506.05865 (2015-06)
LCSTS: A Large Scale Chinese Short Text Summarization Dataset
1509.01626 (2015-09)
Character-level Convolutional Networks for Text Classification *
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 56 more not named here
acc_and_f1 22 other papers
22 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1905.12866 (2019-05)
Semantically Conditioned Dialog Response Generation via Hierarchical Disentangled Self-Attention
1908.09355 (2019-08)
Patient Knowledge Distillation for BERT Model Compression
and 19 more not named here
accuracy 6 other papers
6 dated earlier
2302.02931 (2023-02)
Bitrate-Constrained DRO: Beyond Worst Case Robustness To Unknown Group Shifts
2306.11074 (2023-06)
Simple and Fast Group Robustness by Automatic Feature Reweighting
2309.06717 (2023-09)
Bias Amplification Enhances Minority Group Performance
and 3 more not named here
convert_examples_to_features 5 other papers
5 dated earlier
2010.10418 (2020-10)
CONJNLI: Natural Language Inference Over Conjunctive Sentences
2306.11074 (2023-06)
Simple and Fast Group Robustness by Automatic Feature Reweighting
2309.06717 (2023-09)
Bias Amplification Enhances Minority Group Performance
and 2 more not named here
combine_and_mask 4 other papers
4 dated earlier
2007.04612 (2020-07)
Concept Bottleneck Models
2310.06161 (2023-10)
Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
2405.04715 (2024-05)
Causality Pursuit from Heterogeneous Environments via Neural Adversarial Invariance Learning
and 1 more not named here
crop_and_resize 4 other papers
4 dated earlier
2007.04612 (2020-07)
Concept Bottleneck Models
2310.06161 (2023-10)
Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
2403.03375 (2024-03)
Complexity Matters: Dynamics of Feature Learning in the Presence of Spurious Correlations
and 1 more not named here
get_accs_for_epoch_across_batches 4 other papers
4 dated earlier
2306.11074 (2023-06)
Simple and Fast Group Robustness by Automatic Feature Reweighting
2310.06161 (2023-10)
Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
2312.03044 (2023-12)
REST: Enhancing Group Robustness in DNNs through Reweighted Sparse Training
and 1 more not named here
load_log 4 other papers
4 dated earlier
2306.11074 (2023-06)
Simple and Fast Group Robustness by Automatic Feature Reweighting
2310.06161 (2023-10)
Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
2312.03044 (2023-12)
REST: Enhancing Group Robustness in DNNs through Reweighted Sparse Training
and 1 more not named here
sanitize_df 4 other papers
4 dated earlier
2306.11074 (2023-06)
Simple and Fast Group Robustness by Automatic Feature Reweighting
2310.06161 (2023-10)
Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
2312.03044 (2023-12)
REST: Enhancing Group Robustness in DNNs through Reweighted Sparse Training
and 1 more not named here
get_vgg13 1 other paper
1 dated earlier
2311.16681 (2023-11)
Understanding the (Extra-)Ordinary: Validating Deep Model Decisions with Prototypical Concept-based Explanations
get_vgg16 1 other paper
1 dated earlier
2311.16681 (2023-11)
Understanding the (Extra-)Ordinary: Validating Deep Model Decisions with Prototypical Concept-based Explanations
get_vgg16_bn 1 other paper
1 dated earlier
2311.16681 (2023-11)
Understanding the (Extra-)Ordinary: Validating Deep Model Decisions with Prototypical Concept-based Explanations

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

Deep Neural Networks (DNNs) are increasingly utilized in high-stakes domains like medical diagnostics and autonomous driving where model reliability is critical. However, the research landscape for ensuring this reliability is terminologically fractured across communities that pursue the same goal of ensuring models rely on causally relevant features rather than confounding signals. While frameworks such as distributionally robust optimization (DRO), invariant risk minimization (IRM), shortcut learning, simplicity bias, and the Clever Hans effect all address model failure due to spurious correlations, researchers typically only reference work within their own domains. This reproducibility study unifies these perspectives through a comparative analysis of correction methods under challenging constraints like limited data availability and severe subgroup imbalance. We evaluate recently proposed correction methods based on explainable artificial intelligence (XAI) techniques alongside popular non-XAI baselines using both synthetic and real-world datasets. Findings show that XAI-based methods generally outperform non-XAI approaches, with Counterfactual Knowledge Distillation (CFKD) proving most consistently effective at improving generalization. Our experiments also reveal that the practical application of many methods is hindered by a dependency on group labels, as manual annotation is often infeasible and automated tools like Spectral Relevance Analysis (SpRAy) struggle with complex features and severe imbalance. Furthermore, the scarcity of minority group samples in validation sets renders model selection and hyperparameter tuning unreliable, posing a significant obstacle to the deployment of robust and trustworthy models in safety-critical areas.

For agents

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

get_harvested_code_for_paper("2604.04518")
get_code_for_paper("2604.04518")
have("2604.04518")

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

curl https://syntology.ai/api/ran/2604.04518.json

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

[![Syntology run record](https://syntology.ai/api/ran/2604.04518.svg)](https://syntology.ai/paper/2604.04518)

Connect an agent — have() is free.