SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.03094 · 2023

PopulAtion Parameter Averaging (PAPA)

arXiv · PDF · Open in the Atlas

Code that ran

10 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; 2 failures of our instrument; 1 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/2304.03094.json · badge: /api/ran/2304.03094.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
samsungsailmontreal/papa canonical (classifier label)
classifier basis not recorded on the link
10 ran · 1 no recorded run
FunctionStatusWhere it lives
ResNet18 Ran
ran on a synthesized input
samsungsailmontreal/papa/models/resnet.py
code served (permissive licence) · get_code("d7c3eb24eb759696")
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 ResNet. 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)

ARGS = {
    "model_name": "resnet18",
    "act": "relu",
    "norm": "bn",
    "num_classes": 10,
    "bias": False,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
conv3x3 Ran
ran; our contract draft was wrong, not the code
samsungsailmontreal/papa/models/resnet.py
code served (permissive licence) · get_code("ca893f9498143980")
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 Conv2d. 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)

ARGS = {
    "in_planes": 3,
    "out_planes": 64,
    "stride": 1,
    "bias": False,
}
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 Conv2d — a Conv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv2d 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_blocks Ran
ran on a synthesized input
samsungsailmontreal/papa/models/resnet.py
code served (permissive licence) · get_code("b2174c34ed70a43a")
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 Sequential. 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)

class ResNetBlock(nn.Module):
    def __init__(self, in_channels, out_channels):
        super().__init__()
        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1)
        self.bn = nn.BatchNorm2d(out_channels)
        self.relu = nn.ReLU()
    
    def forward(self, x):
        return self.relu(self.bn(self.conv(x)))

class SimpleResNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 4, kernel_size=7, stride=2, padding=3)
        self.bn1 = nn.BatchNorm2d(4)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        
        self.layer1 = nn.Sequential(
            ResNetBlock(4, 4),
            ResNetBlock(4, 4)
        )
        self.layer2 = nn.Sequential(
            ResNetBlock(4, 8),
            ResNetBlock(8, 8)
        )
        self.layer3 = nn.Sequential(
            ResNetBlock(8, 16),
            ResNetBlock(16, 16)
        )
        self.layer4 = nn.Sequential(
            ResNetBlock(16, 32),
            ResNetBlock(32, 32)
        )
    
    def forward(self, x):
        return x

net = SimpleResNet()

ARGS = {
    "net": net
}
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 Sequential — a Sequential 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
get_layer_perm Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/corr_matching.py
code served (permissive licence) · get_code("af037d161636f160")
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 1, dtype family int, 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 torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Create a simple CNN that outputs spatial feature maps
class SimpleNet(nn.Module):
    def __init__(self, out_channels=4):
        super().__init__()
        self.conv1 = nn.Conv2d(3, out_channels, kernel_size=3, padding=1)
    
    def forward(self, x):
        # Output shape: (batch, channels, height, width)
        return self.conv1(x)

# Create a tiny dataset
class TinyDataset(torch.utils.data.Dataset):
    def __init__(self, num_samples=4):
        self.num_samples = num_samples
    
    def __len__(self):
        return self.num_samples
    
    def __getitem__(self, idx):
        # Return (image, label) tuple
        # Image: (3, 8, 8) - 3 channels, 8x8 spatial
        image = torch.randn(3, 8, 8)
        label = 0
        return image, label

# Build arguments
device = 'cpu'
train_dset = TinyDataset(num_samples=4)
net0 = SimpleNet(out_channels=4)
net1 = SimpleNet(out_channels=4)

ARGS = {
    'device': device,
    'train_dset': train_dset,
    'net0': net0,
    'net1': net1,
    'batch_size': 2,
    'n_iter': 1,
    'shuffle': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family int, 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
get_layer_perm1 Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/corr_matching.py
code served (permissive licence) · get_code("f6cae5f4d5272117")
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 1, dtype family int, 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 torch
import numpy as np
import scipy.optimize

torch.manual_seed(0)

# Create a small correlation matrix
# The function expects a tensor with a .cpu() method and .numpy() conversion
# It performs linear sum assignment which works on 2D arrays
# Using a small 3x3 matrix for testing
corr_mtx = torch.tensor([
    [0.9, 0.1, 0.2],
    [0.3, 0.8, 0.4],
    [0.2, 0.5, 0.7]
], dtype=torch.float32)

ARGS = {
    "corr_mtx": corr_mtx
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family int, 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
make_repaired_net Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/repair.py
code served (permissive licence) · get_code("6cf54613a40deb52")
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 SimpleNet. 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 copy

torch.manual_seed(0)

# Build a small CNN model with Conv2d layers
class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 4, kernel_size=3, padding=1),
            nn.ReLU(),
            nn.Conv2d(4, 8, kernel_size=3, padding=1),
            nn.ReLU(),
        )
        self.classifier = nn.Linear(8, 2)
    
    def forward(self, x):
        x = self.features(x)
        return x

# Create the network
net = SimpleNet()

# Build args object with device attribute
class Args:
    def __init__(self):
        self.device = 'cpu'

args = Args()

ARGS = {
    'net': net,
    'args': args,
}
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 SimpleNet — a SimpleNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleNet 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
make_tracked_net Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/repair.py
code served (permissive licence) · get_code("1601cfedc4402bd6")
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 SimpleNet. 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 copy
import torch
from torch import nn

torch.manual_seed(0)

# Build a small CNN model
class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 16, kernel_size=3, padding=1),
            nn.ReLU(inplace=True),
            nn.Conv2d(16, 32, kernel_size=3, padding=1),
            nn.ReLU(inplace=True),
        )
        self.classifier = nn.Linear(32 * 8 * 8, 10)
    
    def forward(self, x):
        x = self.features(x)
        x = x.view(x.size(0), -1)
        x = self.classifier(x)
        return x

# Build a small args object with device attribute
class Args:
    def __init__(self, device):
        self.device = device

# Create instances
net = SimpleNet()
args = Args(device='cpu')

ARGS = {
    'net': net,
    'args': args,
}
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 SimpleNet — a SimpleNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleNet 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
mix_weights_direct Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/permutations.py
code served (permissive licence) · get_code("d22d36bb8ee5d00b")
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 SimpleNet. 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)

class SimpleNet(nn.Module):
    def __init__(self, input_dim=4, output_dim=3):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, x):
        return self.linear(x)

# Create device (CPU only)
device = 'cpu'

# Create alpha coefficients for weight mixing
alpha = [0.5, 0.3, 0.2]

# Create target network that will receive mixed weights
net = SimpleNet(input_dim=4, output_dim=3)

# Create list of networks to mix
nets = [
    SimpleNet(input_dim=4, output_dim=3),
    SimpleNet(input_dim=4, output_dim=3),
    SimpleNet(input_dim=4, output_dim=3)
]

# Initialize networks with different weights so mixing is meaningful
for i, network in enumerate(nets):
    for param in network.parameters():
        param.data.fill_(float(i + 1))

ARGS = {
    'device': device,
    'alpha': alpha,
    'net': net,
    'nets': nets
}
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 SimpleNet — a SimpleNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleNet 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
resnet20 Ran
ran; our contract draft was wrong, not the code
samsungsailmontreal/papa/models/resnet20.py
code served (permissive licence) · get_code("5ec019fd36dd68e8")
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 ResNet. 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)

ARGS = {
    "num_classes": 10
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
run_corr_matrix Ran
ran on a synthesized input
samsungsailmontreal/papa/recombine/corr_matching.py
code served (permissive licence) · get_code("527355ff9a678a4c")
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 2, dtype family float, shape (4, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple neural network that outputs spatial feature maps
class SimpleNet(nn.Module):
    def __init__(self, out_channels=4):
        super().__init__()
        self.conv = nn.Conv2d(3, out_channels, kernel_size=3, padding=1)
    
    def forward(self, x):
        # Input: (batch, 3, H, W)
        # Output: (batch, out_channels, H, W)
        return self.conv(x)

# Create small dataset with (image, label) tuples
class TinyDataset(torch.utils.data.Dataset):
    def __init__(self, num_samples=4):
        self.num_samples = num_samples
    
    def __len__(self):
        return self.num_samples
    
    def __getitem__(self, idx):
        # Return (image, label)
        image = torch.randn(3, 8, 8)  # 3 channels, 8x8 spatial
        label = torch.tensor(0)
        return image, label

# Instantiate components
device = 'cpu'
train_dset = TinyDataset(num_samples=4)
net0 = SimpleNet(out_channels=4)
net1 = SimpleNet(out_channels=4)

ARGS = {
    'device': device,
    'train_dset': train_dset,
    'net0': net0,
    'net1': net1,
    'batch_size': 2,
    'n_iter': 99999,
    'shuffle': True
}
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 2, dtype family float, shape (4, 4) — shape (4, 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
convert_model Not yet run
unverified (harvested, no recorded run)
samsungsailmontreal/papa/sync_batchnorm/batchnorm.py
code served (permissive licence) · get_code("9cd45e4b4fb9ad28")
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, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 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
conv3x3 8 other papers
5 dated earlier, 3 later
1512.03385 (2015-12)
Deep Residual Learning for Image Recognition
1805.11046 (2018-05)
Scalable Methods for 8-bit Training of Neural Networks
2001.07645 (2020-01)
SAUNet: Shape Attentive U-Net for Interpretable Medical Image Segmentation
and 5 more not named here
convert_model 4 other papers
3 dated earlier, 1 later
1711.07240 (2017-11)
MegDet: A Large Mini-Batch Object Detector
2108.13559 (2021-08)
Music Demixing Challenge 2021
2112.09343 (2021-12)
Domain Adaptation on Point Clouds via Geometry-Aware Implicits
and 1 more not named here
resnet20 1 other paper
1 later
2307.07873 (2023-07)
Why Does Little Robustness Help? A Further Step Towards Understanding Adversarial Transferability

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

Ensemble methods combine the predictions of multiple models to improve performance, but they require significantly higher computation costs at inference time. To avoid these costs, multiple neural networks can be combined into one by averaging their weights. However, this usually performs significantly worse than ensembling. Weight averaging is only beneficial when different enough to benefit from combining them, but similar enough to average well. Based on this idea, we propose PopulAtion Parameter Averaging (PAPA): a method that combines the generality of ensembling with the efficiency of weight averaging. PAPA leverages a population of diverse models (trained on different data orders, augmentations, and regularizations) while slowly pushing the weights of the networks toward the population average of the weights. We also propose PAPA variants (PAPA-all, and PAPA-2) that average weights rarely rather than continuously; all methods increase generalization, but PAPA tends to perform best. PAPA reduces the performance gap between averaging and ensembling, increasing the average accuracy of a population of models by up to 0.8% on CIFAR-10, 1.9% on CIFAR-100, and 1.6% on ImageNet when compared to training independent (non-averaged) models.

For agents

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

get_harvested_code_for_paper("2304.03094")
get_code_for_paper("2304.03094")
have("2304.03094")

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

curl https://syntology.ai/api/ran/2304.03094.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.03094.svg)](https://syntology.ai/paper/2304.03094)

Connect an agent — have() is free.