SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2302.04451 · 2023

Generalization in Graph Neural Networks: Improved PAC-Bayesian Bounds on Graph Diffusion

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 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; 4 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 2 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2302.04451.json · badge: /api/ran/2302.04451.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
neu-statsml-research/generalization-in-graph-neural-networks canonical (classifier label)
classifier basis not recorded on the link
6 ran · 2 no recorded run
virtuosoresearch/generalization-in-graph-neural-networks canonical (classifier label)
classifier basis: classifier_2025
4 ran · 0 no recorded run
FunctionStatusWhere it lives
add_penalty Ran
ran on a synthesized input
neu-statsml-research/generalization-in-graph-neural-networks/src/utils/constraint.py
code served (permissive licence) · get_code("f97515ae0c7bb673")
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 with weight parameters (no bias for simplicity in test)
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3, bias=False)
        self.linear2 = nn.Linear(3, 2, bias=False)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

# Create a state_dict for comparison (optional parameter)
state_dict = {
    'linear1.weight': torch.randn(3, 4),
    'linear2.weight': torch.randn(2, 3),
}

ARGS = {
    'model': model,
    'norm': 'frob',
    '_lambda': 0.1,
    'excluding_key': None,
    'including_key': None,
    'state_dict': state_dict,
}
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
compute_hessians_quantity Ran
ran, honoured the contract we drafted
virtuosoresearch/generalization-in-graph-neural-networks/src/compute_hessian_spectra.py
code served (permissive licence) · get_code("47c29243e21dd172")
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 torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Define a simple model with Linear layers
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
        self.graph_pred_linear = nn.Linear(2, 1)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        x = self.graph_pred_linear(x)
        return x

# Create model and input data
model = SimpleModel()

# Create small input batch (batch_size=2, input_dim=4)
x = torch.randn(2, 4, requires_grad=True)
y = torch.randn(2, 1)

# Forward pass to compute loss
output = model(x)
loss = torch.nn.functional.mse_loss(output, y)

# Create state_dict with the same structure as model weights
# This is used for comparing weights in 'graph_pred_linear' check
state_dict = {
    'linear1.weight': model.linear1.weight.detach().clone(),
    'linear2.weight': model.linear2.weight.detach().clone(),
    'graph_pred_linear.weight': model.graph_pred_linear.weight.detach().clone(),
}

ARGS = {
    'model': model,
    'loss': loss,
    'state_dict': state_dict,
}
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 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
compute_hessians_quantity Ran
ran on a synthesized input
neu-statsml-research/generalization-in-graph-neural-networks/src/compute_hessian_norm.py
code served (permissive licence) · get_code("160e009a6d205294")
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 torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Build a minimal model with Linear and Embedding layers
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = nn.Embedding(num_embeddings=10, embedding_dim=4)
        self.linear1 = nn.Linear(4, 3)
        self.graph_pred_linear = nn.Linear(3, 2)
    
    def forward(self, x):
        # x shape: (batch_size, seq_len)
        embedded = self.embedding(x)  # (batch_size, seq_len, 4)
        # Flatten for linear layer
        flat = embedded.view(embedded.size(0), -1)  # (batch_size, seq_len*4)
        out1 = self.linear1(flat[:, :4])  # (batch_size, 3)
        out2 = self.graph_pred_linear(out1)  # (batch_size, 2)
        return out2

# Create model and input
model = SimpleModel()

# Create a simple loss
input_ids = torch.randint(0, 10, (2, 2))  # batch_size=2, seq_len=2
output = model(input_ids)
target = torch.tensor([[0.5, 1.5], [1.0, 2.0]])
loss = torch.nn.functional.mse_loss(output, target)

# Create state_dict - for layers without 'graph_pred_linear' in name
state_dict = {}
for name, param in model.named_parameters():
    if 'weight' in name:
        state_dict[name] = param.detach().clone()

ARGS = {
    "model": model,
    "loss": loss,
    "state_dict": state_dict
}
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 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
compute_hessians_trace Ran
ran, honoured the contract we drafted
virtuosoresearch/generalization-in-graph-neural-networks/src/compute_hessian_spectra.py
code served (permissive licence) · get_code("a869544c3ef6df3b")
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 (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Build a small model with Linear layers
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(4, 3)
        self.fc2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.fc1(x)
        x = self.fc2(x)
        return x

# Create model instance
model = SimpleModel()

# Create dummy input and target
x = torch.randn(2, 4)
target = torch.randint(0, 2, (2,))

# Forward pass to get loss
output = model(x)
loss = nn.CrossEntropyLoss()(output, target)

ARGS = {
    "model": model,
    "loss": loss,
    "maxIter": 5,
    "tol": 1e-6
}
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 ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
deep_copy Ran
ran on a synthesized input
neu-statsml-research/generalization-in-graph-neural-networks/src/utils/constraint.py
code served (permissive licence) · get_code("d2ee7320cf52c3fc")
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

torch.manual_seed(0)

# The function deep_copy takes a state_dict and clones/detaches each tensor value
# state_dict is expected to be a dict where values are tensors with .clone() and .detach() methods

state_dict = {
    "layer1.weight": torch.randn(4, 3),
    "layer1.bias": torch.randn(2),
    "layer2.weight": torch.randn(5, 4),
}

ARGS = {
    "state_dict": state_dict
}
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 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
eval Ran
ran, honoured the contract we drafted
virtuosoresearch/generalization-in-graph-neural-networks/src/finetune.py
code served (permissive licence) · get_code("c9ab4e1ae3be9759")
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
import torch.nn as nn
from torch_geometric.data import Data, DataLoader

torch.manual_seed(0)

class SimpleGNNModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x, edge_index, edge_attr, batch):
        # Return logits with shape (num_nodes, num_tasks)
        return self.linear(x)

# Create a simple GNN model
model = SimpleGNNModel()

# Create sample batched graph data
# Each batch item needs: x, edge_index, edge_attr, batch, y
def create_sample_batch():
    # Node features: 2 nodes, 4 features
    x = torch.randn(2, 4)
    # Edge index: 2 edges (COO format)
    edge_index = torch.tensor([[0, 1], [1, 0]], dtype=torch.long).t().contiguous()
    # Edge attributes: 2 edges, 3 features
    edge_attr = torch.randn(2, 3)
    # Batch assignment
    batch = torch.tensor([0, 0], dtype=torch.long)
    # Target labels: 2 nodes, 3 tasks (with values -1, 0, or 1)
    y = torch.tensor([[-1, 1, -1], [1, -1, 1]], dtype=torch.float32)
    
    data = Data(x=x, edge_index=edge_index, edge_attr=edge_attr, batch=batch, y=y)
    return data

# Create loader with 2 batches
batch1 = create_sample_batch()
batch2 = create_sample_batch()
loader = [batch1, batch2]

device = 'cpu'

ARGS = {
    'args': None,
    'model': model,
    'device': device,
    'loader': loader
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, 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 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_weights Ran
ran; our contract draft was wrong, not the code
virtuosoresearch/generalization-in-graph-neural-networks/src/compute_hessian_spectra.py
code served (permissive licence) · get_code("7144f95dffb38b67")
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
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
        self.relu = nn.ReLU()
        self.linear3 = nn.Linear(2, 1)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.relu(x)
        x = self.linear2(x)
        x = self.linear3(x)
        return x

model = SimpleModel()

ARGS = {
    "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 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
get_weights Ran
ran on a synthesized input
neu-statsml-research/generalization-in-graph-neural-networks/src/compute_hessian_norm.py
code served (permissive licence) · get_code("7f274d9139984ff5")
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
import torch.nn as nn

torch.manual_seed(0)

# Build a simple model with Linear and Embedding layers
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.embedding = nn.Embedding(10, 5)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        return x

model = SimpleModel()

ARGS = {
    "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 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
graph_data_obj_to_nx_simple Ran
ran, honoured the contract we drafted
neu-statsml-research/generalization-in-graph-neural-networks/src/loader.py
code served (permissive licence) · get_code("287484b87212f47d")
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 Graph. 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 torch_geometric.data import Data

torch.manual_seed(0)

# Create a small graph Data object with 3 atoms and 2 bonds
# Atom features: [atomic_num_idx, chirality_tag_idx]
atom_features = torch.tensor([
    [6, 0],      # atom 0: carbon, no chirality
    [7, 1],      # atom 1: nitrogen, chirality 1
    [8, 0]       # atom 2: oxygen, no chirality
], dtype=torch.float32)

# Edge index: [2, 4] tensor representing edges
# Each pair of rows represents one undirected edge (stored twice in PyG convention)
edge_index = torch.tensor([
    [0, 1, 1, 0],    # source atoms
    [1, 0, 2, 2]     # target atoms
], dtype=torch.long)

# Edge attributes: [bond_type_idx, bond_dir_idx]
edge_attr = torch.tensor([
    [1, 0],          # bond 0-1: single bond, no direction
    [1, 0],          # bond 1-0: single bond, no direction
    [2, 1],          # bond 1-2: double bond, direction 1
    [2, 1]           # bond 2-1: double bond, direction 1
], dtype=torch.float32)

# Create PyG Data object
data_obj = Data(x=atom_features, edge_index=edge_index, edge_attr=edge_attr)

ARGS = {
    "data": data_obj
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx, numpy, torch, torch_geometric — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Graph — a Graph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Graph output the recorded ndim 1 and dtype family int 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
reset_idxes Ran
ran; our contract draft was wrong, not the code
neu-statsml-research/generalization-in-graph-neural-networks/src/util.py
code served (permissive licence) · get_code("6203c137185fa051")
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 networkx as nx

# Create a small graph with non-sequential node indices
G = nx.Graph()
G.add_edges_from([(5, 10), (10, 15), (15, 5)])
G.add_node(20)

ARGS = {
    "G": G
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx — 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_hessians_trace Not yet run
unverified (harvested, no recorded run)
neu-statsml-research/generalization-in-graph-neural-networks/src/compute_hessian_norm.py
code served (permissive licence) · get_code("b354514b745d77fc")
How it ran

no invocation recorded.

get_test_acc Not yet run
unverified (harvested, no recorded run)
neu-statsml-research/generalization-in-graph-neural-networks/src/parse_result.py
pointer only (licence: MIT); the stored copy get_code serves, from LiuChuang0059/StructMAE, is pointer only (licence: NONE) · get_code("7a2bb855a3e02bfe")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from LiuChuang0059/StructMAE, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 5 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 2 later) and 6 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
graph_data_obj_to_nx_simple 9 other papers
1 dated earlier, 8 later
1905.12265 (2019-05)
Strategies for Pre-training Graph Neural Networks
2306.07812 (2023-06)
Automated 3D Pre-Training for Molecular Property Prediction
2309.01788 (2023-09)
Hierarchical Grammar-Induced Geometry for Data-Efficient Molecular Property Prediction
and 6 more not named here
get_test_acc 2 other papers
2 later
2309.01788 (2023-09)
Hierarchical Grammar-Induced Geometry for Data-Efficient Molecular Property Prediction
2404.15806 (2024-04)
Where to Mask: Structure-Guided Masking for Graph Masked Autoencoders
add_penalty 1 other paper
1 dated earlier
2111.04578 (2021-11)
Improved Regularization and Robustness for Fine-tuning in Neural Networks
deep_copy 1 other paper
1 dated earlier
2111.04578 (2021-11)
Improved Regularization and Robustness for Fine-tuning in Neural Networks
reset_idxes 1 other paper
1 dated earlier
1905.12265 (2019-05)
Strategies for Pre-training Graph Neural 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 2026-11-22.

Repositories linked to this paper

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

Abstract

Graph neural networks are widely used tools for graph prediction tasks. Motivated by their empirical performance, prior works have developed generalization bounds for graph neural networks, which scale with graph structures in terms of the maximum degree. In this paper, we present generalization bounds that instead scale with the largest singular value of the graph neural network's feature diffusion matrix. These bounds are numerically much smaller than prior bounds for real-world graphs. We also construct a lower bound of the generalization gap that matches our upper bound asymptotically. To achieve these results, we analyze a unified model that includes prior works' settings (i.e., convolutional and message-passing networks) and new settings (i.e., graph isomorphism networks). Our key idea is to measure the stability of graph neural networks against noise perturbations using Hessians. Empirically, we find that Hessian-based measurements correlate with the observed generalization gaps of graph neural networks accurately. Optimizing noise stability properties for fine-tuning pretrained graph neural networks also improves test performance on several graph-level classification tasks.

For agents

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

get_harvested_code_for_paper("2302.04451")
get_code_for_paper("2302.04451")
have("2302.04451")

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

curl https://syntology.ai/api/ran/2302.04451.json

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

[![Syntology run record](https://syntology.ai/api/ran/2302.04451.svg)](https://syntology.ai/paper/2302.04451)

Connect an agent — have() is free.