SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2003.02228 · 2020

PushNet: Efficient and Adaptive Neural Message Passing

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 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 7 have no contract verdict; 0 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/2003.02228.json · badge: /api/ran/2003.02228.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
buschju/pushnet canonical (classifier label)
classifier basis: census
7 ran · 0 no recorded run
FunctionStatusWhere it lives
evaluate_model Ran
ran on a synthesized input
buschju/pushnet/src/training.py
code served (permissive licence) · get_code("03b3243eaff8a95c")
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 dict. 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
from torch_geometric.data import Data

torch.manual_seed(0)

# Create a minimal model that satisfies the function's usage
class MinimalGNNModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)  # 4 input features, 3 output classes
        self.l2_reg_params = [self.linear.weight, self.linear.bias]
    
    def forward(self, data):
        # Return log probabilities (shape: [num_nodes, num_classes])
        return torch.nn.functional.log_softmax(self.linear(data.x), dim=1)

# Create model
model = MinimalGNNModel()

# Create a minimal Data object with the required attributes
num_nodes = 6
num_features = 4
num_classes = 3

data = Data(
    x=torch.randn(num_nodes, num_features),
    y=torch.randint(0, num_classes, (num_nodes,)),
    train_mask=torch.tensor([True, True, False, False, False, False]),
    val_mask=torch.tensor([False, False, True, True, False, False]),
    test_mask=torch.tensor([False, False, False, False, True, True])
)

# Define l2_reg parameter
l2_reg = 0.01

ARGS = {
    'model': model,
    'data': data,
    'l2_reg': l2_reg
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torch_geometric — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
flatten_dictionary Ran
ran on a synthesized input
buschju/pushnet/src/utils.py
code served (permissive licence) · get_code("613d60256d9e04fc")
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 dict. 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
from typing import Dict, Any, Optional

torch.manual_seed(0)

# Analyze the function:
# - dictionary: Dict[str, Any] - needs to be a nested dict with string keys
# - separator: str - default is '.'
# - prefix: Optional[str] - default is ''
# The function recursively flattens nested dictionaries, using the separator
# to join keys and optionally prefix them.

# Create a nested dictionary to test the flattening
nested_dict = {
    'level1_a': {
        'level2_a': 'value_a',
        'level2_b': 42
    },
    'level1_b': {
        'level2_c': 3.14,
        'level2_d': {
            'level3_a': 'deep_value'
        }
    },
    'simple_key': 'simple_value'
}

ARGS = {
    'dictionary': nested_dict,
    'separator': '.',
    'prefix': ''
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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_preprocessed_path Ran
ran on a synthesized input
buschju/pushnet/src/data_utils.py
code served (permissive licence) · get_code("ca86a40ede8d9a7f")
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 os
import tempfile

# Create a temporary directory for testing
temp_dir = tempfile.mkdtemp()

ARGS = {
    'dataset_name': 'mnist',
    'data_root': temp_dir,
    'sub_folder': 'train',
    'make_dirs': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
load_adj Ran
ran on a synthesized input
buschju/pushnet/src/data_utils.py
code served (permissive licence) · get_code("3e95ac42238a2d04")
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 os
import torch
import tempfile
import shutil

torch.manual_seed(0)

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
data_root = temp_dir

# Create preprocessed directory structure
dataset_name = "test_dataset"
preprocessed_path = os.path.join(data_root, 'preprocessed', dataset_name)
os.makedirs(preprocessed_path, exist_ok=True)

# Create dummy edge_index tensor (2 x num_edges format)
edge_index = torch.tensor([[0, 1, 2], [1, 2, 0]], dtype=torch.long)

# Save the edge_index file with different naming conventions
# We need to create files that match the possible naming patterns
file_name = dataset_name
torch.save(edge_index, os.path.join(preprocessed_path, f'{file_name}.edge_index'))

# Also create a version with all flags to ensure the function can find at least one file
file_name_full = f'{dataset_name}_largest_cc_self_loops_sym'
torch.save(edge_index, os.path.join(preprocessed_path, f'{file_name_full}.edge_index'))

# Create edge_weight for the full version
edge_weight = torch.tensor([0.5, 0.7, 0.3], dtype=torch.float)
torch.save(edge_weight, os.path.join(preprocessed_path, f'{file_name_full}.edge_weight'))

ARGS = {
    'only_largest_cc': False,
    'add_self_loops': False,
    'adj_normalization': None,
    'dataset_name': dataset_name,
    'data_root': data_root
}
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
load_ppr Ran
ran on a synthesized input
buschju/pushnet/src/data_utils.py
code served (permissive licence) · get_code("dba42869fc139e4a")
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 os
import tempfile
import torch
from torch import Tensor

torch.manual_seed(0)

# Create a temporary directory to serve as data_root
temp_dir = tempfile.mkdtemp()

# Create the preprocessed/ppr directory structure with dummy files
dataset_name = "test_dataset"
ppr_dir = os.path.join(temp_dir, 'preprocessed', dataset_name, 'ppr')
os.makedirs(ppr_dir, exist_ok=True)

# Create dummy edge_index and edge_weight tensors
edge_index = torch.tensor([[0, 1, 2], [1, 2, 0]], dtype=torch.long)
edge_weight = torch.tensor([0.5, 0.3, 0.2], dtype=torch.float)

# Save the tensors with the expected filename pattern
file_name = f'{dataset_name}_largest_cc_self_loops_symmetric_alpha=1.00e-01_epsilon=1.00e-04'
torch.save(edge_index, os.path.join(ppr_dir, file_name + '.edge_index'))
torch.save(edge_weight, os.path.join(ppr_dir, file_name + '.edge_weight'))

ARGS = {
    'only_largest_cc': True,
    'add_self_loops': True,
    'adj_normalization': 'symmetric',
    'alpha': 0.1,
    'epsilon': 1e-4,
    'dataset_name': dataset_name,
    'data_root': temp_dir,
}
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
train_random_splits Ran
ran on a synthesized input
buschju/pushnet/src/training.py
code served (permissive licence) · get_code("45d23b9fb662e6cd")
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 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 torch
import torch.nn as nn
from torch_geometric.data import Data
from torch import Tensor

torch.manual_seed(0)

# Create a minimal model that satisfies the function's usage
class SimpleGNNModel(nn.Module):
    def __init__(self, input_dim: int, output_dim: int):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
        self.l2_reg_params = [self.linear.weight]
    
    def forward(self, data: Data) -> Tensor:
        # Return log probabilities with shape [num_nodes, num_classes]
        return torch.log_softmax(self.linear(data.x), dim=1)
    
    def reset_parameters(self):
        self.linear.reset_parameters()

# Create synthetic data
num_nodes = 10
input_dim = 4
num_classes = 3

x = torch.randn(num_nodes, input_dim)
y = torch.randint(0, num_classes, (num_nodes,))

# Create train/val/test masks for default split
train_mask = torch.tensor([True, True, False, False, False, False, False, False, False, False])
val_mask = torch.tensor([False, False, True, True, False, False, False, False, False, False])
test_mask = torch.tensor([False, False, False, False, True, True, True, True, True, True])

data = Data(x=x, y=y)
data.train_mask = train_mask
data.val_mask = val_mask
data.test_mask = test_mask

# Create model
model = SimpleGNNModel(input_dim=input_dim, output_dim=num_classes)

# Create random splits - list of dicts with different train/val/test masks
splits = []
for split_idx in range(2):
    # Generate different random splits
    indices = torch.randperm(num_nodes)
    train_size = 3
    val_size = 2
    
    train_indices = indices[:train_size]
    val_indices = indices[train_size:train_size + val_size]
    test_indices = indices[train_size + val_size:]
    
    split_train_mask = torch.zeros(num_nodes, dtype=torch.bool)
    split_val_mask = torch.zeros(num_nodes, dtype=torch.bool)
    split_test_mask = torch.zeros(num_nodes, dtype=torch.bool)
    
    split_train_mask[train_indices] = True
    split_val_mask[val_indices] = True
    split_test_mask[test_indices] = True
    
    splits.append({
        'train_mask': split_train_mask,
        'val_mask': split_val_mask,
        'test_mask': split_test_mask,
    })

ARGS = {
    'splits': splits,
    'num_runs_per_split': 1,
    'model': model,
    'data': data,
    'learning_rate': 0.01,
    'patience': 2,
    'max_epochs': 5,
    'l2_reg': 0.0001,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torch_geometric — 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
train_single_split Ran
ran on a synthesized input
buschju/pushnet/src/training.py
code served (permissive licence) · get_code("4da72336053bb605")
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 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 torch
import torch.nn as nn
from torch_geometric.data import Data

torch.manual_seed(0)

# Create a minimal GNN model that satisfies the function's requirements
class MinimalGNN(nn.Module):
    def __init__(self, num_nodes=10, num_features=4, num_classes=3):
        super().__init__()
        self.lin = nn.Linear(num_features, num_classes)
        self.l2_reg_params = [self.lin.weight, self.lin.bias]
    
    def forward(self, data):
        # Return log probabilities for all nodes
        return torch.nn.functional.log_softmax(self.lin(data.x), dim=1)
    
    def reset_parameters(self):
        self.lin.reset_parameters()

# Create a minimal Data object with all required attributes
num_nodes = 10
num_features = 4
num_classes = 3

x = torch.randn(num_nodes, num_features)
y = torch.randint(0, num_classes, (num_nodes,))

# Create train/val/test masks
train_mask = torch.zeros(num_nodes, dtype=torch.bool)
val_mask = torch.zeros(num_nodes, dtype=torch.bool)
test_mask = torch.zeros(num_nodes, dtype=torch.bool)

train_mask[:4] = True
val_mask[4:7] = True
test_mask[7:] = True

data = Data(x=x, y=y)
data.train_mask = train_mask
data.val_mask = val_mask
data.test_mask = test_mask

# Create model instance
model = MinimalGNN(num_nodes=num_nodes, num_features=num_features, num_classes=num_classes)

# Define ARGS dictionary with all required parameters
ARGS = {
    'num_runs': 2,
    'model': model,
    'data': data,
    'learning_rate': 0.01,
    'patience': 3,
    'max_epochs': 5,
    'l2_reg': 0.0001
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torch_geometric — 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

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 7 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 2026-11-21.

Repositories linked to this paper

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

Abstract

Message passing neural networks have recently evolved into a state-of-the-art approach to representation learning on graphs. Existing methods perform synchronous message passing along all edges in multiple subsequent rounds and consequently suffer from various shortcomings: Propagation schemes are inflexible since they are restricted to $k$-hop neighborhoods and insensitive to actual demands of information propagation. Further, long-range dependencies cannot be modeled adequately and learned representations are based on correlations of fixed locality. These issues prevent existing methods from reaching their full potential in terms of prediction performance. Instead, we consider a novel asynchronous message passing approach where information is pushed only along the most relevant edges until convergence. Our proposed algorithm can equivalently be formulated as a single synchronous message passing iteration using a suitable neighborhood function, thus sharing the advantages of existing methods while addressing their central issues. The resulting neural network utilizes a node-adaptive receptive field derived from meaningful sparse node neighborhoods. In addition, by learning and combining node representations over differently sized neighborhoods, our model is able to capture correlations on multiple scales. We further propose variants of our base model with different inductive bias. Empirical results are provided for semi-supervised node classification on five real-world datasets following a rigorous evaluation protocol. We find that our models outperform competitors on all datasets in terms of accuracy with statistical significance. In some cases, our models additionally provide faster runtime.

For agents

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

get_harvested_code_for_paper("2003.02228")
get_code_for_paper("2003.02228")
have("2003.02228")

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

curl https://syntology.ai/api/ran/2003.02228.json

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

[![Syntology run record](https://syntology.ai/api/ran/2003.02228.svg)](https://syntology.ai/paper/2003.02228)

Connect an agent — have() is free.