SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1912.03250 · 2019

Differentially Private Synthetic Mixed-Type Data Generation For Unsupervised Learning

arXiv · PDF · Open in the Atlas

Code that ran

11 of 11 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; 3 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/1912.03250.json · badge: /api/ran/1912.03250.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
DPautoGAN/DPautoGAN canonical (classifier label)
classifier basis: census
11 ran · 0 no recorded run
FunctionStatusWhere it lives
balance Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/evaluation.py
code served (permissive licence) · get_code("f3c0636a6c2e8bf2")
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 numpy as np
from sklearn.model_selection import train_test_split

# Create sample data with imbalanced labels
np.random.seed(0)

# Create data: 10 samples with 4 features each
data = np.random.randn(10, 4).astype(np.float32)

# Create imbalanced labels: 7 samples with label 0, 3 samples with label 1
label = np.array([0, 0, 0, 0, 0, 0, 0, 1, 1, 1])

ARGS = {
    "data": data,
    "label": label,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
compute_rdp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
DPautoGAN/DPautoGAN/mimic/rdp_accountant.py
code served (permissive licence) · get_code("dfcf05f8070a1a7d")
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 ndarray, ndim 1, dtype family float, shape (4,) 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
import math
from scipy import special

# Based on the function signature and implementation, we need to provide:
# - q: sampling rate (float between 0 and 1)
# - noise_multiplier: ratio of std dev to l2-sensitivity (float)
# - steps: number of steps (int)
# - orders: RDP orders (scalar or array)

ARGS = {
    'q': 0.01,  # sampling rate between 0 and 1
    'noise_multiplier': 1.5,  # noise multiplier (sigma)
    'steps': 100,  # number of steps
    'orders': np.array([1.5, 2.0, 3.0, 5.0])  # array of RDP orders
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, six — 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 (4,) — shape (4,) 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_rdp_from_ledger Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/rdp_accountant.py
code served (permissive licence) · get_code("1bd33e484e06a1e7")
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 ndarray, ndim 1, dtype family float, shape (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 math
import numpy as np
from scipy import special
import six

# Define a small stand-in class for Query objects
class Query:
    def __init__(self, noise_stddev, l2_norm_bound):
        self.noise_stddev = noise_stddev
        self.l2_norm_bound = l2_norm_bound

# Define a small stand-in class for Sample objects
class Sample:
    def __init__(self, selection_probability, queries):
        self.selection_probability = selection_probability
        self.queries = queries

# Create a minimal ledger with sample data
# Each sample has a selection_probability and a list of queries
# Each query has noise_stddev and l2_norm_bound
query1 = Query(noise_stddev=0.5, l2_norm_bound=1.0)
query2 = Query(noise_stddev=0.7, l2_norm_bound=2.0)
sample1 = Sample(selection_probability=0.1, queries=[query1, query2])

query3 = Query(noise_stddev=0.6, l2_norm_bound=1.5)
sample2 = Sample(selection_probability=0.2, queries=[query3])

ledger = [sample1, sample2]

# Define orders as both scalar and array variants
orders = np.array([1.5, 2.0, 3.0])

ARGS = {
    'ledger': ledger,
    'orders': orders
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, six — 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 (3,) — shape (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
convert_to_3digit_icd9 Ran
ran; our contract draft was wrong, not the code
DPautoGAN/DPautoGAN/mimic/mimic_dataset.py
code served (permissive licence) · get_code("9f1a1575f6901240")
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

def convert_to_3digit_icd9(dxStr):
    if dxStr.startswith('E'):
        if len(dxStr) > 4:
            return dxStr[:4]
        else:
            return dxStr
    elif len(dxStr) > 3:
        return dxStr[:3]
    else:
        return dxStr

ARGS = {
    'dxStr': 'E1234'
}
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
convert_to_icd9 Ran
ran; our contract draft was wrong, not the code
DPautoGAN/DPautoGAN/mimic/mimic_dataset.py
code served (permissive licence) · get_code("f782bcdaf42568da")
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

ARGS = {
    'dxStr': 'E1234'
}
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 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_data_loaders Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/sampling.py
code served (permissive licence) · get_code("d18aa2d6a3614265")
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 torch
from torch.utils.data import TensorDataset
import numpy as np

torch.manual_seed(0)

# Create a small dataset with 10 samples, each with 4 features and 1 label
X = torch.randn(10, 4)
y = torch.randint(0, 2, (10,))
dataset = TensorDataset(X, y)

# Define arguments for get_data_loaders
ARGS = {
    'minibatch_size': 3,
    'microbatch_size': 2,
    'iterations': 2
}
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 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_patient_matrix Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/mimic_dataset.py
code served (permissive licence) · get_code("cb6574d99c975563")
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 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 tempfile
import os
from datetime import datetime

# Create temporary files with admission and diagnosis data
admission_data = """ROW_ID,SUBJECT_ID,HADM_ID,ADMITTIME,DISCHTIME,ADMISSION_TYPE,ADMISSION_LOCATION,DISCHARGE_LOCATION,INSURANCE,LANGUAGE,MARITAL_STATUS,ETHNICITY,EDREGTIME,EDOUTTIME,DIAGNOSIS
1,100,1000,2020-01-15 10:30:00,2020-01-20 14:00:00,EMERGENCY,EMERGENCY ROOM ADMITTER,HOME HEALTH CARE,Medicare,ENGL,MARRIED,WHITE,,,
2,100,1001,2020-03-10 08:15:00,2020-03-15 09:30:00,URGENT,CLINIC REFERRAL,LONG TERM CARE,Medicare,ENGL,MARRIED,WHITE,,,
3,101,1002,2020-02-05 12:00:00,2020-02-10 16:45:00,EMERGENCY,EMERGENCY ROOM ADMITTER,HOME,Medicaid,ENGL,SINGLE,ASIAN,,,
4,101,1003,2020-04-20 14:20:00,2020-04-25 11:10:00,URGENT,TRANSFER FROM HOSP,HOME HEALTH CARE,Medicaid,ENGL,SINGLE,ASIAN,,,
"""

diagnosis_data = """ROW_ID,SUBJECT_ID,HADM_ID,SEQ_NUM,ICD9_CODE,SHORT_TITLE,LONG_TITLE
1,100,1000,1,"E119",Type 2 diabetes without complications,
2,100,1000,2,"I10",Essential (primary) hypertension,
3,100,1001,1,"E119",Type 2 diabetes without complications,
4,100,1001,2,"J441",Chronic obstructive pulmonary disease with (acute) exacerbation,
5,101,1002,1,"I10",Essential (primary) hypertension,
6,101,1002,2,"J441",Chronic obstructive pulmonary disease with (acute) exacerbation,
7,101,1003,1,"E119",Type 2 diabetes without complications,
8,101,1003,2,"I10",Essential (primary) hypertension,
"""

# Write to temporary files
temp_dir = tempfile.gettempdir()
admission_file = os.path.join(temp_dir, "admissions_test.csv")
diagnosis_file = os.path.join(temp_dir, "diagnoses_test.csv")

with open(admission_file, 'w') as f:
    f.write(admission_data)

with open(diagnosis_file, 'w') as f:
    f.write(diagnosis_data)

ARGS = {
    'admissionFile': admission_file,
    'diagnosisFile': diagnosis_file,
    'binary_count': 'binary'
}
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 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
get_prediction_score Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/evaluation.py
code served (permissive licence) · get_code("8d151a57c7d4a1bf")
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

# Define a simple model class that has fit and predict_proba methods
class SimpleModel:
    def __init__(self):
        self.model = LogisticRegression(random_state=0, max_iter=1000)
    
    def fit(self, X, y):
        self.model.fit(X, y)
        return self
    
    def predict_proba(self, X):
        return self.model.predict_proba(X)

# Create small training and test data
np.random.seed(0)
X_train = np.random.randn(10, 4)  # 10 samples, 4 features
y_train = np.array([0, 1, 0, 1, 0, 1, 0, 1, 0, 1])

X_test = np.random.randn(5, 4)  # 5 samples, 4 features
y_test = np.array([0, 1, 0, 1, 0])

# Define a simple scoring function
def score_fn(y_true, y_pred):
    from sklearn.metrics import roc_auc_score
    return roc_auc_score(y_true, y_pred)

ARGS = {
    "X_train": X_train,
    "y_train": y_train,
    "X_test": X_test,
    "y_test": y_test,
    "Model": SimpleModel,
    "score": score_fn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
get_privacy_spent Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/rdp_accountant.py
code served (permissive licence) · get_code("690416000476a525")
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 math

# Construct arguments for get_privacy_spent
orders = np.array([1.5, 2.0, 3.0, 5.0])
rdp = np.array([0.1, 0.15, 0.2, 0.25])
target_eps = 0.5
target_delta = None

ARGS = {
    "orders": orders,
    "rdp": rdp,
    "target_eps": target_eps,
    "target_delta": target_delta,
}
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 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
isolate_column Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/evaluation.py
code served (permissive licence) · get_code("18e3ea94db9eb751")
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 pandas as pd
import numpy as np

# Create a small dataframe for testing
data = {
    'feature_1': [1.0, 2.0],
    'feature_2': [3.0, 4.0],
    'feature_3': [5.0, 6.0],
    'target': [7.0, 8.0]
}
df = pd.DataFrame(data)

# Define the arguments
ARGS = {
    'df': df,
    'column': 'target'
}
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 — 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
make_optimizer_class Ran
ran on a synthesized input
DPautoGAN/DPautoGAN/mimic/dp_optimizer.py
code served (permissive licence) · get_code("02dc1abf1033d768")
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 type. 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
import torch.optim as optim

torch.manual_seed(0)

# Create a simple model with parameters
class SimpleModel(nn.Module):
    def __init__(self):
        super(SimpleModel, self).__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = torch.relu(self.linear1(x))
        x = self.linear2(x)
        return x

model = SimpleModel()

# Create parameters for the optimizer
# We use SGD as the base optimizer class
optimizer_cls = optim.SGD

# Create parameter groups (what SGD expects)
params = model.parameters()

ARGS = {
    'cls': optimizer_cls
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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 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, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 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.

FunctionAlso present inEarliest-datedOthers named
compute_rdp 1 other paper
1 later
2402.00205 (2024-02)
Decentralised, Collaborative, and Privacy-preserving Machine Learning for Multi-Hospital Data
convert_to_3digit_icd9 1 other paper
1 dated earlier
1703.06490 (2017-03)
Generating Multi-label Discrete Patient Records using Generative Adversarial Networks
convert_to_icd9 1 other paper
1 dated earlier
1703.06490 (2017-03)
Generating Multi-label Discrete Patient Records using Generative Adversarial Networks

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We introduce the DP-auto-GAN framework for synthetic data generation, which combines the low dimensional representation of autoencoders with the flexibility of Generative Adversarial Networks (GANs). This framework can be used to take in raw sensitive data and privately train a model for generating synthetic data that will satisfy similar statistical properties as the original data. This learned model can generate an arbitrary amount of synthetic data, which can then be freely shared due to the post-processing guarantee of differential privacy. Our framework is applicable to unlabeled mixed-type data, that may include binary, categorical, and real-valued data. We implement this framework on both binary data (MIMIC-III) and mixed-type data (ADULT), and compare its performance with existing private algorithms on metrics in unsupervised settings. We also introduce a new quantitative metric able to detect diversity, or lack thereof, of synthetic data.

For agents

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

get_harvested_code_for_paper("1912.03250")
get_code_for_paper("1912.03250")
have("1912.03250")

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

curl https://syntology.ai/api/ran/1912.03250.json

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

[![Syntology run record](https://syntology.ai/api/ran/1912.03250.svg)](https://syntology.ai/paper/1912.03250)

Connect an agent — have() is free.