SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.09263 · ICML · 2025

Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence

Lei Shi, Yuankai Luo, Xiao-Ming Wu, Buaa Cn>

arXiv · PDF · Open in the Atlas

Code that ran

7 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; 1 honoured the contract we drafted, 2 violated it and 0 have no contract verdict; 4 failures of our instrument; 5 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2502.09263.json · badge: /api/ran/2502.09263.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
LUOyk1999/GNNPlus canonical (classifier label)
classifier basis not recorded on the link
7 ran · 5 no recorded run
FunctionStatusWhere it lives
accuracy_SBM Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
LUOyk1999/GNNPlus/GNNPlus/logger.py
code served (permissive licence) · get_code("ecf14287d3465c48")
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 float64, ndim 0, 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 torch

torch.manual_seed(0)

# Create mock targets as a torch tensor
# The function expects targets to have .cpu().detach().numpy() method
targets = torch.tensor([0, 1, 0, 1, 2, 2], dtype=torch.long)

# Create mock predictions as integers (class indices)
# Should match targets shape
pred_int = np.array([0, 1, 0, 1, 2, 1], dtype=np.int64)

ARGS = {
    "targets": targets,
    "pred_int": pred_int
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
eval_spearmanr Ran
ran; our contract draft was wrong, not the code
LUOyk1999/GNNPlus/GNNPlus/logger.py
code served (permissive licence) · get_code("9dc78001d467fc9d")
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 numpy as np
from scipy.stats import stats

# Test case 1: 1D arrays (single task)
y_true_1d = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
y_pred_1d = np.array([1.1, 2.2, 2.9, 4.1, 4.8])

# Test case 2: 2D arrays (multiple tasks) with some NaN values
y_true_2d = np.array([
    [1.0, 2.0, 3.0],
    [2.0, np.nan, 4.0],
    [3.0, 3.0, np.nan],
    [4.0, 4.0, 5.0],
    [5.0, 5.0, 6.0]
], dtype=float)

y_pred_2d = np.array([
    [1.1, 2.1, 3.1],
    [2.1, 1.9, 4.1],
    [3.1, 3.1, 5.1],
    [4.1, 3.9, 5.1],
    [5.1, 5.1, 6.1]
], dtype=float)

# Use 1D case for ARGS
ARGS = {
    'y_true': y_true_1d,
    'y_pred': y_pred_1d
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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_final_pretrained_ckpt Ran
ran; our contract draft was wrong, not the code
LUOyk1999/GNNPlus/GNNPlus/finetuning.py
code served (permissive licence) · get_code("1bb331bf70a78136")
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
import shutil

# Create a temporary directory with checkpoint files
temp_dir = tempfile.mkdtemp()

# Create some checkpoint files with epoch numbers
checkpoint_files = ['1.ckpt', '2.ckpt', '5.ckpt', '3.ckpt']
for ckpt_file in checkpoint_files:
    filepath = os.path.join(temp_dir, ckpt_file)
    with open(filepath, 'w') as f:
        f.write('dummy checkpoint data')

# Define arguments for the function
ARGS = {
    'ckpt_dir': temp_dir
}
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
is_seed Ran
ran, violated the contract we drafted
LUOyk1999/GNNPlus/GNNPlus/agg_runs.py
pointer only (licence: MIT); the stored copy get_code serves, from sunjss/over-aggregating, is pointer only (licence: NONE) · get_code("c98a702be0660ac9")
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 sunjss/over-aggregating, 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. A run is recorded.

is_split Ran
ran, violated the contract we drafted
LUOyk1999/GNNPlus/GNNPlus/agg_runs.py
pointer only (licence: MIT); the stored copy get_code serves, from sunjss/over-aggregating, is pointer only (licence: NONE) · get_code("438635c08adfe2aa")
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 sunjss/over-aggregating, 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. A run is recorded.

join_list Ran
ran, honoured the contract we drafted
LUOyk1999/GNNPlus/GNNPlus/agg_runs.py
pointer only (licence: MIT); the stored copy get_code serves, from sunjss/over-aggregating, is pointer only (licence: NONE) · get_code("40e98e1ecbf39f34")
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 sunjss/over-aggregating, 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. A run is recorded.

subtoken_cross_entropy Ran
ran; our contract draft was wrong, not the code
LUOyk1999/GNNPlus/GNNPlus/loss/subtoken_prediction_loss.py
code served (permissive licence) · get_code("47e492109e6d308f")
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
from torch_geometric.graphgym.config import cfg

torch.manual_seed(0)

# Configure the global config object for subtoken_prediction task
cfg.dataset.task_type = 'subtoken_prediction'
cfg.model.loss_fun = 'cross_entropy'

# Create pred_list: a list of tensors with logits for each subtoken position
# Each tensor should have shape (batch_size, num_classes)
batch_size = 2
num_classes = 4
num_subtokens = 3

pred_list = [
    torch.randn(batch_size, num_classes),
    torch.randn(batch_size, num_classes),
    torch.randn(batch_size, num_classes)
]

# Create true: a dict with 'y_arr' containing target token indices for each position
# Shape should be (batch_size, num_subtokens)
true = {
    'y_arr': torch.randint(0, num_classes, (batch_size, num_subtokens))
}

ARGS = {
    'pred_list': pred_list,
    'true': true
}
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 — 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
init_model_from_pretrained Not yet run
unverified (harvested, no recorded run)
LUOyk1999/GNNPlus/GNNPlus/finetuning.py
code served (permissive licence) · get_code("3db4474abe13fdee")
How it ran

no invocation recorded.

l1_losses Not yet run
unverified (harvested, no recorded run)
LUOyk1999/GNNPlus/GNNPlus/loss/l1.py
code served (permissive licence) · get_code("d59c374db2cf64ae")
How it ran

no invocation recorded.

load_pretrained_model_cfg Not yet run
unverified (harvested, no recorded run)
LUOyk1999/GNNPlus/GNNPlus/finetuning.py
code served (permissive licence) · get_code("d796cdea5889c847")
How it ran

no invocation recorded.

multilabel_cross_entropy Not yet run
unverified (harvested, no recorded run)
LUOyk1999/GNNPlus/GNNPlus/loss/multilabel_classification_loss.py
code served (permissive licence) · get_code("f1cbbbc857d8b871")
How it ran

no invocation recorded.

weighted_cross_entropy Not yet run
unverified (harvested, no recorded run)
LUOyk1999/GNNPlus/GNNPlus/loss/weighted_cross_entropy.py
code served (permissive licence) · get_code("4632e6a524984022")
How it ran

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 12 distinct functions with an identity, 12 are also present in other papers whose code we hold (12 with at least one paper dated earlier, 0 the same month, 9 later) and 0 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
get_final_pretrained_ckpt 27 other papers
25 dated earlier, 2 later
2302.04181 (2023-02)
Attending to Graph Transformers
2302.08647 (2023-02)
Multiresolution Graph Transformers and Wavelet Positional Encoding for Learning Hierarchical Structures
2305.08018 (2023-05)
DRew: Dynamically Rewired Message Passing with Delay
and 24 more not named here
is_seed 15 other papers
14 dated earlier, 1 later
2302.04181 (2023-02)
Attending to Graph Transformers
2309.00367 (2023-09)
Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark
2310.19285 (2023-10)
Facilitating Graph Neural Networks with Random Walk on Simplicial Complexes
and 12 more not named here
is_split 15 other papers
14 dated earlier, 1 later
2302.04181 (2023-02)
Attending to Graph Transformers
2309.00367 (2023-09)
Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark
2310.19285 (2023-10)
Facilitating Graph Neural Networks with Random Walk on Simplicial Complexes
and 12 more not named here
join_list 15 other papers
14 dated earlier, 1 later
2302.04181 (2023-02)
Attending to Graph Transformers
2309.00367 (2023-09)
Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark
2310.19285 (2023-10)
Facilitating Graph Neural Networks with Random Walk on Simplicial Complexes
and 12 more not named here
init_model_from_pretrained 10 other papers
10 dated earlier
2302.04181 (2023-02)
Attending to Graph Transformers
2309.00367 (2023-09)
Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
and 7 more not named here
load_pretrained_model_cfg 9 other papers
9 dated earlier
2302.04181 (2023-02)
Attending to Graph Transformers
2309.00367 (2023-09)
Where Did the Gap Go? Reassessing the Long-Range Graph Benchmark
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
and 6 more not named here
eval_spearmanr 6 other papers
5 dated earlier, 1 later
2305.17589 (2023-05)
Graph Inductive Biases in Transformers without Message Passing
2404.13604 (2024-04)
CKGConv: General Graph Convolution with Continuous Kernels
2405.13806 (2024-05)
A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition
and 3 more not named here
accuracy_SBM 5 other papers
4 dated earlier, 1 later
2305.17589 (2023-05)
Graph Inductive Biases in Transformers without Message Passing
2404.13604 (2024-04)
CKGConv: General Graph Convolution with Continuous Kernels
2405.13806 (2024-05)
A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition
and 2 more not named here
l1_losses 5 other papers
4 dated earlier, 1 later
2305.17589 (2023-05)
Graph Inductive Biases in Transformers without Message Passing
2404.13604 (2024-04)
CKGConv: General Graph Convolution with Continuous Kernels
2405.13806 (2024-05)
A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition
and 2 more not named here
multilabel_cross_entropy 5 other papers
4 dated earlier, 1 later
2305.17589 (2023-05)
Graph Inductive Biases in Transformers without Message Passing
2404.13604 (2024-04)
CKGConv: General Graph Convolution with Continuous Kernels
2405.13806 (2024-05)
A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition
and 2 more not named here
subtoken_cross_entropy 3 other papers
2 dated earlier, 1 later
2305.17589 (2023-05)
Graph Inductive Biases in Transformers without Message Passing
2404.13604 (2024-04)
CKGConv: General Graph Convolution with Continuous Kernels
2606.14259 (2026-06)
Beyond a Single Explanation of the Adam-SGD Gap
weighted_cross_entropy 1 other paper
1 dated earlier
2411.16278 (2024-11)
Even Sparser Graph Transformers

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

Message-passing Graph Neural Networks (GNNs) are often criticized for their limited expressiveness, issues like over-smoothing and oversquashing, and challenges in capturing long-range dependencies. Conversely, Graph Transformers (GTs) are regarded as superior due to their employment of global attention mechanisms, which potentially mitigate these challenges. Literature frequently suggests that GTs outperform GNNs in graph-level tasks, especially for graph classification and regression on small molecular graphs. In this study, we explore the untapped potential of GNNs through an enhanced framework, GNN + , which integrates six widely used techniques: edge feature integration, normalization, dropout, residual connections, feed-forward networks, and positional encoding, to effectively tackle graph-level tasks. We conduct a systematic re-evaluation of three classic GNNs-GCN, GIN, and GatedGCN-enhanced by the GNN + framework across 14 well-known graph-level datasets. Our results reveal that, contrary to prevailing beliefs, these classic GNNs consistently match or surpass the performance of GTs, securing topthree rankings across all datasets and achieving first place in eight. Furthermore, they demonstrate greater efficiency, running several times faster than GTs on many datasets. This highlights the potential of simple GNN architectures, challenging the notion that complex mechanisms in GTs are essential for superior graph-level performance. Our source code is available at https://github.com/LUOyk1999/GNNPlus.

For agents

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

get_harvested_code_for_paper("2502.09263")
get_code_for_paper("2502.09263")
have("2502.09263")

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

curl https://syntology.ai/api/ran/2502.09263.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.09263.svg)](https://syntology.ai/paper/2502.09263)

Connect an agent — have() is free.