SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.14960 · ICLR · 2022

Domino: Discovering Systematic Errors with Cross-Modal Embeddings

Jared Dunnmon, Christopher Ré, James Zou, Khaled Saab, Christopher Lee-Messer, Sabri Eyuboglu, Maya Varma, Jean-Benoit Delbrouck

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 6 failures of our instrument; 2 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/2203.14960.json · badge: /api/ran/2203.14960.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
gregdeon/spotlight unrelated (classifier label)
classifier basis: classifier_2025
5 ran · 0 no recorded run
hazyresearch/domino marked official in Papers with Code
Papers with Code archive
4 ran · 2 no recorded run
FunctionStatusWhere it lives
ResetOnPlateau Ran
ran on a synthesized input · constructed an object; no result was computed
gregdeon/spotlight/torch_spotlight/spotlight.py
pointer only (licence: NONE) · get_code("f3b9828cefd56507")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gregdeon/spotlight 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.

Variable Ran
ran on a synthesized input
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("cc7e8d94a77d5498")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned ndarray, 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

class Variable:

    def __init__(self, name: str, weight: float=1.0):
        self.name = name
        self.weight = weight

    def preprocess(self, value: np.ndarray):
        return value

    def check(self, value: np.ndarray, n_components: int):
        return value


INIT_ARGS = {
    "name": "test_variable",
    "weight": 2.5
}

ARGS = {
    "value": np.array([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]]),
    "n_components": 3
}

METHOD = "check"

var = Variable(**INIT_ARGS)
result = getattr(var, METHOD)(**ARGS)
print(f"Variable name: {var.name}")
print(f"Variable weight: {var.weight}")
print(f"Result shape: {result.shape}")
print(f"Result:\n{result}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
_check_X Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("51ffaa11557d91a6")
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 (4, 3) 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.utils import check_array

# Create a small valid input array
X = np.array([[1.0, 2.0, 3.0],
              [4.0, 5.0, 6.0],
              [7.0, 8.0, 9.0],
              [10.0, 11.0, 12.0]], dtype=np.float64)

ARGS = {
    'X': X,
    'n_components': 2,
    'n_features': 3,
    'ensure_min_samples': 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, sklearn — 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 (4, 3) — shape (4, 3) 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
_compute_log_det_cholesky Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("5f267401600f3c19")
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 1, dtype family float, shape (2,) 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: 'full' covariance type
n_components_full = 2
n_features_full = 4
matrix_chol_full = np.random.rand(n_components_full, n_features_full, n_features_full)

# Test case 2: 'tied' covariance type
n_features_tied = 4
matrix_chol_tied = np.random.rand(n_features_tied, n_features_tied)

# Test case 3: 'diag' covariance type
n_components_diag = 2
n_features_diag = 4
matrix_chol_diag = np.random.rand(n_components_diag, n_features_diag)

# Test case 4: 'spherical' covariance type
n_components_spherical = 2
n_features_spherical = 4
matrix_chol_spherical = np.random.rand(n_components_spherical)

# Use 'full' as the primary test case
ARGS = {
    'matrix_chol': matrix_chol_full,
    'covariance_type': 'full',
    'n_features': n_features_full
}
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 1, dtype family float, shape (2,) — shape (2,) 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
_estimate_log_gaussian_prob Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("2f6f28561fa49c7c")
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 (2, 3) 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

# Set random seed for reproducibility
np.random.seed(0)

# Define parameters
n_samples = 2
n_features = 4
n_components = 3

# Create sample data X: (n_samples, n_features)
X = np.random.randn(n_samples, n_features).astype(np.float64)

# Create means: (n_components, n_features)
means = np.random.randn(n_components, n_features).astype(np.float64)

# Choose covariance type
covariance_type = 'full'

# Create precisions_chol based on covariance_type
if covariance_type == 'full':
    # Shape: (n_components, n_features, n_features)
    precisions_chol = np.random.randn(n_components, n_features, n_features).astype(np.float64)
    # Make it upper triangular (Cholesky form)
    precisions_chol = np.triu(precisions_chol)
    # Ensure positive diagonal for valid Cholesky
    for i in range(n_components):
        for j in range(n_features):
            precisions_chol[i, j, j] = abs(precisions_chol[i, j, j]) + 0.1
elif covariance_type == 'tied':
    # Shape: (n_features, n_features)
    precisions_chol = np.random.randn(n_features, n_features).astype(np.float64)
    precisions_chol = np.triu(precisions_chol)
    for j in range(n_features):
        precisions_chol[j, j] = abs(precisions_chol[j, j]) + 0.1
elif covariance_type == 'diag':
    # Shape: (n_components, n_features)
    precisions_chol = np.abs(np.random.randn(n_components, n_features).astype(np.float64)) + 0.1
elif covariance_type == 'spherical':
    # Shape: (n_components,)
    precisions_chol = np.abs(np.random.randn(n_components).astype(np.float64)) + 0.1

ARGS = {
    'X': X,
    'means': means,
    'precisions_chol': precisions_chol,
    'covariance_type': covariance_type,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, sklearn — 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, 3) — shape (2, 3) 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
md_adversary_weights Ran
ran; our contract draft was wrong, not the code
gregdeon/spotlight/torch_spotlight/spotlight.py
pointer only (licence: NONE) · get_code("1c0ef4dae3ce7c5d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gregdeon/spotlight 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.

md_objective Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gregdeon/spotlight/torch_spotlight/spotlight.py
pointer only (licence: NONE) · get_code("8e89a99c6cd34993")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gregdeon/spotlight 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.

run_spotlight Ran
ran; our contract draft was wrong, not the code
gregdeon/spotlight/torch_spotlight/spotlight.py
pointer only (licence: NONE) · get_code("428763fcb8b5a166")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gregdeon/spotlight 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.

squared_distance_kernel Ran
ran, honoured the contract we drafted
gregdeon/spotlight/torch_spotlight/spotlight.py
pointer only (licence: NONE) · get_code("53db6c58e4b86d70")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gregdeon/spotlight 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.

BaseMixture Not yet run
unverified (harvested, no recorded run)
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("b43fe52491a8c7f5")
How it ran

no invocation recorded.

GeneralizedMixture Not yet run
unverified (harvested, no recorded run)
hazyresearch/domino/domino/_slice/mixture/_generalized.py
code served (permissive licence) · get_code("0bfab385e191b9ac")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 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.

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 that achieve high overall accuracy often make systematic errors on important subsets (or slices) of data. Identifying underperforming slices is particularly challenging when working with high-dimensional inputs (e.g. images, audio), where important slices are often unlabeled. In order to address this issue, recent studies have proposed automated slice discovery methods (SDMs), which leverage learned model representations to mine input data for slices on which a model performs poorly. To be useful to a practitioner, these methods must identify slices that are both underperforming and coherent (i.e. united by a human-understandable concept). However, no quantitative evaluation framework currently exists for rigorously assessing SDMs with respect to these criteria. Additionally, prior qualitative evaluations have shown that SDMs often identify slices that are incoherent. In this work, we address these challenges by first designing a principled evaluation framework that enables a quantitative comparison of SDMs across 1,235 slice discovery settings in three input domains (natural images, medical images, and time-series data). Then, motivated by the recent development of powerful cross-modal representation learning approaches, we present Domino, an SDM that leverages cross-modal embeddings and a novel error-aware mixture model to discover and describe coherent slices. We find that Domino accurately identifies 36% of the 1,235 slices in our framework -a 12 percentage point improvement over prior methods. Further, Domino is the first SDM that can provide natural language descriptions of identified slices, correctly generating the exact name of the slice in 35% of settings.

For agents

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

get_harvested_code_for_paper("2203.14960")
get_code_for_paper("2203.14960")
have("2203.14960")

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

curl https://syntology.ai/api/ran/2203.14960.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.14960.svg)](https://syntology.ai/paper/2203.14960)

Connect an agent — have() is free.