SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.04567 · 2026

Generative Modeling under Non-Monotone MAR Missingness via Approximate Wasserstein Gradient Flows

Johannes Lederer, Jeffrey Näf, Gitte Kremling

arXiv · PDF · Open in the Atlas

Code that ran

8 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; 0 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 7 failures of our instrument; 3 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/2604.04567.json · badge: /api/ran/2604.04567.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
yujhml/MIRI-Imputation community
Papers with Code archive
5 ran · 2 no recorded run
gkremling/FLOWGEM community
Papers with Code archive
3 ran · 1 no recorded run
FunctionStatusWhere it lives
MINE Ran
ran; our contract draft was wrong, not the code
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("f2bdb58b90f28871")
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. 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 tensors with distinct dimensions
# x0: batch_size=2, feature_dim=4
# y0: batch_size=2, feature_dim=3
x0 = torch.randn(2, 4)
y0 = torch.randn(2, 3)

ARGS = {
    'x0': x0,
    'y0': y0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, tqdm — 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
MLP Ran
ran on a synthesized input · constructed an object; no result was computed
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("e96bb7e0947a8068")
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 an instance of its own class, MLP: the run constructed the object and did not call it, so no result was computed. 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 MLP.__init__ takes two parameters: dx and dy
# These represent input dimension and output dimension (or context dimension)
# Looking at the forward method, it concatenates x with something of size dy,
# so dx is the feature dimension and dy is another dimension

dx = 4
dy = 3

ARGS = {
    'dx': dx,
    'dy': dy,
}
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 MLP — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MLP 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
evaluate_sigma Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gkremling/FLOWGEM/flowgem.py
code served (permissive licence) · get_code("39fb66daeeb5a51a")
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. 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, 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

torch.manual_seed(0)

# Create small tensors with distinct dimensions
# xpi: shape (n_pi, d) - samples from pi distribution
# xrho: shape (n_rho, d) - samples from rho distribution
# d = 4 (feature dimension)

d = 4
n_pi = 3
n_rho = 3

xpi = torch.randn(n_pi, d, dtype=torch.float64)
xrho = torch.randn(n_rho, d, dtype=torch.float64)

# sigma: bandwidth parameter for RBF kernel
sigma = 1.0

# splits_pi and splits_rho: lists of (train_indices, val_indices) tuples
# We need to create cross-validation splits
# For simplicity, use 2 splits with train/val indices

splits_pi = [
    (torch.tensor([0, 1]), torch.tensor([2])),
    (torch.tensor([2]), torch.tensor([0, 1])),
]

splits_rho = [
    (torch.tensor([0, 1]), torch.tensor([2])),
    (torch.tensor([2]), torch.tensor([0, 1])),
]

ARGS = {
    'xpi': xpi,
    'xrho': xrho,
    'sigma': sigma,
    'splits_pi': splits_pi,
    'splits_rho': splits_rho,
}
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 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
loss Ran
ran; our contract draft was wrong, not the code
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("5bc00938c66e0254")
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 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

torch.manual_seed(0)

# Create a simple model that takes concatenated [x, y] as input
class SimpleModel(nn.Module):
    def __init__(self, input_dim, output_dim=1):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, z):
        return self.linear(z)

# Set up dimensions
batch_size = 2
x_dim = 4
y_dim = 3

# Create tensors
x = torch.randn(batch_size, x_dim)
y = torch.randn(batch_size, y_dim)

# Create model that expects concatenated input of size x_dim + y_dim
model = SimpleModel(input_dim=x_dim + y_dim, output_dim=1)

ARGS = {
    "x": x,
    "y": y,
    "model": model
}
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 Tensor, 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
mmd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("d4d24272a4c0a1b3")
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 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

torch.manual_seed(0)

# Create two small tensors for X and Y
# X: shape (2, 4) - 2 samples, 4 dimensions
# Y: shape (3, 4) - 3 samples, 4 dimensions
X = torch.randn(2, 4, dtype=torch.float32)
Y = torch.randn(3, 4, dtype=torch.float32)

ARGS = {
    'X': X,
    'Y': Y,
    'kernel': 'rbf',
    'sigma': None
}
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 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
opt_wb Ran
ran; our contract draft was wrong, not the code
gkremling/FLOWGEM/flowgem.py
code served (permissive licence) · get_code("f7fc8ec6b79b1741")
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. 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 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)

# Tiny tensors for testing
n = 2  # number of samples
d = 4  # feature dimension
n_pi = 3  # number of pi points
n_rho = 2  # number of rho points

# Create small sample data
x = torch.randn(n, d, dtype=torch.float64)
xpi = torch.randn(n_pi, d, dtype=torch.float64)
xrho = torch.randn(n_rho, d, dtype=torch.float64)
sigma = torch.tensor(1.0, dtype=torch.float64)

ARGS = {
    'xpi': xpi,
    'xrho': xrho,
    'x': x,
    'sigma': sigma,
    'block_size': 1000
}
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 tuple — a tuple 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
sigma_heuristic_pairs Ran
ran; our contract draft was wrong, not the code
gkremling/FLOWGEM/flowgem.py
code served (permissive licence) · get_code("7f9301da834c201b")
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

torch.manual_seed(0)

# Construct a small tensor X with shape (n, d)
# The function accesses X[i] and X[j], computes norms along dim=1
# So we need a 2D tensor with at least 2 rows and some feature dimension
X = torch.randn(8, 4)  # 8 samples, 4 features

# num_pairs controls how many random pairs to sample
num_pairs = 1000

ARGS = {
    "X": X,
    "num_pairs": num_pairs,
}
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 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
train_test_split Ran
ran; our contract draft was wrong, not the code
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("515a6390da9f8f4b")
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. 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)

# Create small tensors for x and y
# x: (batch_size, feature_dim) = (10, 4)
# y: (batch_size, output_dim) = (10, 3)
x = torch.randn(10, 4)
y = torch.randn(10, 3)

ARGS = {
    "x": x,
    "y": y,
}
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 tuple — a tuple 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
rectified_impute Not yet run
unverified (harvested, no recorded run)
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("c9c2fd75adcabf96")
How it ran

no invocation recorded.

sample_flowgem Not yet run
unverified (harvested, no recorded run)
gkremling/FLOWGEM/flowgem.py
code served (permissive licence) · get_code("1dd274ffb60d63fa")
How it ran

no invocation recorded.

sample_ode Not yet run
unverified (harvested, no recorded run)
yujhml/MIRI-Imputation/src/imputer.py
code served (permissive licence) · get_code("8719c6af909c358e")
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

The prevalence of missing values in data science poses a substantial risk to any further analyses. Despite a wealth of research, principled nonparametric methods to deal with general non-monotone missingness are still scarce. Instead, ad-hoc imputation methods are often used, for which it remains unclear whether the correct distribution can be recovered. In this paper, we propose FLOWGEM, a principled iterative method for generating a complete dataset from a dataset with values Missing at Random (MAR). Motivated by convergence results of the ignoring maximum likelihood estimator, our approach minimizes the expected Kullback-Leibler (KL) divergence between the observed data distribution and the distribution of the generated sample over different missingness patterns. To minimize the KL divergence, we employ a discretized particle evolution of the corresponding Wasserstein Gradient Flow, where the velocity field is approximated using a local linear estimator of the density ratio. This construction yields a data generation scheme that iteratively transports an initial particle ensemble toward the target distribution. Simulation studies and real-data benchmarks demonstrate that FLOWGEM achieves state-of-the-art performance across a range of settings, including the challenging case of non-monotone MAR mechanisms. Together, these results position FLOWGEM as a principled and practical alternative to existing imputation methods, and a decisive step towards closing the gap between theoretical rigor and empirical performance.

For agents

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

get_harvested_code_for_paper("2604.04567")
get_code_for_paper("2604.04567")
have("2604.04567")

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

curl https://syntology.ai/api/ran/2604.04567.json

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

[![Syntology run record](https://syntology.ai/api/ran/2604.04567.svg)](https://syntology.ai/paper/2604.04567)

Connect an agent — have() is free.