SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2210.10769 · ICML · 2023

"Why did the Model Fail?": Attributing Model Performance Changes to Distribution Shifts

Shalmali Joshi, Marzyeh Ghassemi, Haoran Zhang, Harvineet Singh

arXiv · PDF · Open in the Atlas

Code that ran

5 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; 1 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 1 failure of our instrument; 3 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/2210.10769.json · badge: /api/ran/2210.10769.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
mlforhealth/expl_perf_drop canonical (classifier label)
classifier basis not recorded on the link
5 ran · 3 no recorded run
FunctionStatusWhere it lives
AUROC Ran
ran on a synthesized input
mlforhealth/expl_perf_drop/expl_perf_drop/metrics.py
code served (permissive licence) · get_code("ed5bdae1388ade09")
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 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 torch
import numpy as np
import pandas as pd
from sklearn.base import BaseEstimator, ClassifierMixin

torch.manual_seed(0)

class DummyClassifier(BaseEstimator, ClassifierMixin):
    """Minimal classifier with predict_proba method."""
    def __init__(self):
        pass
    
    def predict_proba(self, X):
        """Return dummy probabilities for binary classification."""
        n_samples = X.shape[0]
        proba = np.random.rand(n_samples, 2)
        proba = proba / proba.sum(axis=1, keepdims=True)
        return proba

# Create a simple dataframe with features and target
np.random.seed(0)
data = pd.DataFrame({
    'feature_1': np.random.randn(10),
    'feature_2': np.random.randn(10),
    'feature_3': np.random.randn(10),
    'feature_4': np.random.randn(10),
    'Y': np.random.randint(0, 2, 10)
})

model = DummyClassifier()
subset_cols = ['feature_1', 'feature_2', 'feature_3']
weight = np.ones(10)
target_name = 'Y'

ARGS = {
    'model': model,
    'data': data,
    'subset_cols': subset_cols,
    'weight': weight,
    'target_name': target_name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn, torch — 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
accuracy Ran
ran on a synthesized input
mlforhealth/expl_perf_drop/expl_perf_drop/metrics.py
code served (permissive licence) · get_code("4b746b056a598558")
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 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 torch
import numpy as np
import pandas as pd
from sklearn.base import BaseEstimator, ClassifierMixin

torch.manual_seed(0)

# Create a minimal model that has a predict method
class MinimalClassifier(BaseEstimator, ClassifierMixin):
    def predict(self, X):
        # Return binary predictions based on input shape
        n_samples = len(X) if hasattr(X, '__len__') else X.shape[0]
        return np.random.randint(0, 2, n_samples)

# Create sample data as a pandas DataFrame
# The function expects data[subset_cols] and data[target_name] to work
data = pd.DataFrame({
    'feature_1': np.array([0.1, 0.2, 0.3, 0.4]),
    'feature_2': np.array([0.5, 0.6, 0.7, 0.8]),
    'feature_3': np.array([0.9, 1.0, 1.1, 1.2]),
    'feature_4': np.array([1.3, 1.4, 1.5, 1.6]),
    'Y': np.array([0, 1, 1, 0])
})

model = MinimalClassifier()

ARGS = {
    'model': model,
    'data': data,
    'subset_cols': ['feature_1', 'feature_2', 'feature_3'],
    'weight': None,
    'target_name': 'Y'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn, torch — 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
brier Ran
ran on a synthesized input
mlforhealth/expl_perf_drop/expl_perf_drop/metrics.py
code served (permissive licence) · get_code("f408cb23a303554c")
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 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 numpy as np
from sklearn.linear_model import LogisticRegression
import pandas as pd

# Create a simple model with predict_proba method
model = LogisticRegression()

# Create training data for the model
X_train = np.array([[1.0, 2.0, 3.0, 4.0],
                     [2.0, 3.0, 4.0, 5.0],
                     [3.0, 4.0, 5.0, 6.0],
                     [4.0, 5.0, 6.0, 7.0]])
y_train = np.array([0, 1, 0, 1])

# Fit the model
model.fit(X_train, y_train)

# Create a pandas DataFrame with data
data = pd.DataFrame({
    'X1': [1.0, 2.0],
    'X2': [2.0, 3.0],
    'X3': [3.0, 4.0],
    'X4': [4.0, 5.0],
    'Y': [0, 1]
})

# Define subset_cols to test the subset_cols parameter
subset_cols = ['X1', 'X2', 'X3', 'X4']

# Define sample weights
weight = np.array([1.0, 1.0])

# Define target name
target_name = 'Y'

ARGS = {
    'model': model,
    'data': data,
    'subset_cols': subset_cols,
    'weight': weight,
    'target_name': target_name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn — 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
combinations_base Ran
ran; our contract draft was wrong, not the code
mlforhealth/expl_perf_drop/expl_perf_drop/experiments.py
pointer only (licence: MIT); the stored copy get_code serves, from hzhang0/auc_bias, is pointer only (licence: NONE) · get_code("5a9f7480b832c15a")
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 hzhang0/auc_bias, 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.

get_slurm_jobs Ran
ran, honoured the contract we drafted
mlforhealth/expl_perf_drop/expl_perf_drop/launchers.py
code served (permissive licence) · get_code("f2059a864ab0dc16")
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 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 subprocess
import sys

def get_slurm_jobs(user):
    out = subprocess.run(['squeue -u ' + user], shell=True, stdout=subprocess.PIPE).stdout.decode(sys.stdout.encoding)
    a = list(filter(lambda x: len(x) > 0, map(lambda x: x.split(), out.split('\n'))))
    queued, running = ([], [])
    for i in a:
        if i[0].isnumeric():
            if i[4].strip() == 'PD':
                queued.append(int(i[0]))
            else:
                running.append(int(i[0]))
    return (queued, running)

ARGS = {
    'user': 'testuser'
}
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 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
block_until_running Not yet run
unverified (harvested, no recorded run)
mlforhealth/expl_perf_drop/expl_perf_drop/launchers.py
code served (permissive licence) · get_code("bc6dc62bdfa1e0d9")
How it ran

no invocation recorded.

combinations Not yet run
unverified (harvested, no recorded run)
mlforhealth/expl_perf_drop/expl_perf_drop/experiments.py
pointer only (licence: MIT); the stored copy get_code serves, from hzhang0/auc_bias, is pointer only (licence: NONE) · get_code("f64641d62dc45aa9")
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 hzhang0/auc_bias, 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. No invocation recorded.

get_hparams Not yet run
unverified (harvested, no recorded run)
mlforhealth/expl_perf_drop/expl_perf_drop/experiments.py
pointer only (licence: MIT); the stored copy get_code serves, from hzhang0/auc_bias, is pointer only (licence: NONE) · get_code("73d0cb5bbc5c4f0c")
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 hzhang0/auc_bias, 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. No invocation recorded.

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, 5 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 5 later) and 3 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
block_until_running 2 other papers
2 later
2302.12254 (2023-02)
Change is Hard: A Closer Look at Subpopulation Shift
2401.06091 (2024-01)
A Closer Look at AUROC and AUPRC under Class Imbalance
get_slurm_jobs 2 other papers
2 later
2302.12254 (2023-02)
Change is Hard: A Closer Look at Subpopulation Shift
2401.06091 (2024-01)
A Closer Look at AUROC and AUPRC under Class Imbalance
combinations 1 other paper
1 later
2401.06091 (2024-01)
A Closer Look at AUROC and AUPRC under Class Imbalance
combinations_base 1 other paper
1 later
2401.06091 (2024-01)
A Closer Look at AUROC and AUPRC under Class Imbalance
get_hparams 1 other paper
1 later
2401.06091 (2024-01)
A Closer Look at AUROC and AUPRC under Class Imbalance

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 models frequently experience performance drops under distribution shifts. The underlying cause of such shifts may be multiple simultaneous factors such as changes in data quality, differences in specific covariate distributions, or changes in the relationship between label and features. When a model does fail during deployment, attributing performance change to these factors is critical for the model developer to identify the root cause and take mitigating actions. In this work, we introduce the problem of attributing performance differences between environments to distribution shifts in the underlying data generating mechanisms. We formulate the problem as a cooperative game where the players are distributions. We define the value of a set of distributions to be the change in model performance when only this set of distributions has changed between environments, and derive an importance weighting method for computing the value of an arbitrary set of distributions. The contribution of each distribution to the total performance change is then quantified as its Shapley value. We demonstrate the correctness and utility of our method on synthetic, semi-synthetic, and real-world case studies, showing its effectiveness in attributing performance changes to a wide range of distribution shifts.

For agents

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

get_harvested_code_for_paper("2210.10769")
get_code_for_paper("2210.10769")
have("2210.10769")

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

curl https://syntology.ai/api/ran/2210.10769.json

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

[![Syntology run record](https://syntology.ai/api/ran/2210.10769.svg)](https://syntology.ai/paper/2210.10769)

Connect an agent — have() is free.