SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.20829 · 2026

Beyond the Academic Monoculture: A Unified Framework and Industrial Perspective for Attributed Graph Clustering

Yue Liu, Stan Li, Xinwang Liu, Yongchao Liu, Yunhui Liu, Tao Zheng

arXiv · PDF · Open in the Atlas

Code that ran

4 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 4 have no contract verdict; 0 failures of our instrument; 3 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/2603.20829.json · badge: /api/ran/2603.20829.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
Cloudy1225/PyAGC community
Papers with Code archive
4 ran · 3 no recorded run
FunctionStatusWhere it lives
LargeS3GCEncoder Ran
ran on a synthesized input
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("49aef555b6f83383")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (2, 8) 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
from torch import Tensor

class LargeS3GCEncoder(nn.Module):
    """
    Encoder for large-scale graphs (e.g., ogbn-papers100M).

    Uses 2-layer MLPs for better expressiveness.
    Unlike Medium encoder, this doesn't store node embeddings internally -
    they are managed externally by the S3GC model for memory efficiency.

    Architecture:
        :math:`\\bar{X} = \\text{PReLU}(W_2 \\cdot \\text{PReLU}(W_1 \\tilde{A}X)) +
        \\text{PReLU}(W_4 \\cdot \\text{PReLU}(W_3 S_kX)) + I`

    where I (identity embeddings) are passed as an argument rather than stored.
    """

    def __init__(self, in_channels: int, hidden_channels: int):
        super().__init__()
        self.w1 = nn.Linear(in_channels, hidden_channels, bias=True)
        self.w2 = nn.Linear(hidden_channels, hidden_channels, bias=True)
        self.w3 = nn.Linear(in_channels, hidden_channels, bias=True)
        self.w4 = nn.Linear(hidden_channels, hidden_channels, bias=True)
        self.prelu1 = nn.PReLU(hidden_channels)
        self.prelu2 = nn.PReLU(hidden_channels)
        self.reset_parameters()

    def reset_parameters(self):
        self.w1.bias.data.fill_(0.0)
        self.w2.bias.data.fill_(0.0)
        self.w3.bias.data.fill_(0.0)
        self.w4.bias.data.fill_(0.0)

    def forward(self, AX: Tensor, SX: Tensor, iden: Tensor) -> Tensor:
        """
        Forward pass with explicit iden parameter.

        Args:
            AX (Tensor): Pre-multiplied features :math:`\\tilde{A}X` of shape
                :obj:`(batch_size, in_channels)`.
            SX (Tensor): Diffusion features :math:`S_kX` of shape
                :obj:`(batch_size, in_channels)`.
            iden (Tensor): Learnable node embeddings of shape
                :obj:`(batch_size, hidden_channels)`.

        Returns:
            Normalized embeddings of shape :obj:`(batch_size, hidden_channels)`.
        """
        import torch.nn.functional as F
        return F.normalize(self.w2(self.prelu1(self.w1(AX))) + self.w4(self.prelu2(self.w3(SX))) + iden, p=2, dim=1)


INIT_ARGS = {
    "in_channels": 4,
    "hidden_channels": 8,
}

ARGS = {
    "AX": torch.randn(2, 4),
    "SX": torch.randn(2, 4),
    "iden": torch.randn(2, 8),
}

if __name__ == "__main__":
    encoder = LargeS3GCEncoder(**INIT_ARGS)
    output = encoder(**ARGS)
    print("Output shape:", output.shape)
    print("Output:", output)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 (2, 8) — shape (2, 8) 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
LossOutput Ran
ran on a synthesized input
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("7cc27d12a569b359")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import Tensor
from dataclasses import dataclass
from typing import Dict

@dataclass
class LossOutput:
    """
    Unified loss output format for training.

    This class encapsulates both the total loss (used for backpropagation)
    and individual loss components (used for logging and monitoring).

    Args:
        total (Tensor): The total loss scalar for backpropagation.
        components (Dict[str, float]): Dictionary of individual loss components
            for logging purposes, e.g., :obj:`{'reconstruction': 0.5, 'kl': 0.3}`.

    Example:
        >>> loss_output = LossOutput(
        ...     total=torch.tensor(1.5),
        ...     components={'ali': 0.8, 'nei': 0.5, 'spa': 0.2}
        ... )
        >>> print(loss_output.log_string("Epoch 01: "))
        Epoch 01: Loss: 1.5000, ALI: 0.8000, NEI: 0.5000, SPA: 0.2000
    """
    total: Tensor
    components: Dict[str, float]

    def __float__(self) -> float:
        """Enables :obj:`float(loss_output)` for compatibility."""
        return float(self.total.item())

    def log_string(self, prefix: str='') -> str:
        """
        Generates a formatted log string for printing.

        Args:
            prefix (str, optional): Prefix string to prepend. (default: :obj:`""`)

        Returns:
            Formatted string like :obj:`"Loss: 1.50, ALI: 0.80, NEI: 0.50"`.
        """
        parts = [f'Loss: {self.total.item():.4f}']
        for name, value in self.components.items():
            parts.append(f'{name.upper()}: {value:.4f}')
        return f"{prefix}{', '.join(parts)}"


INIT_ARGS = {
    'total': torch.tensor(1.5),
    'components': {'ali': 0.8, 'nei': 0.5, 'spa': 0.2}
}

ARGS = {
    'prefix': 'Epoch 01: '
}

METHOD = 'log_string'

if __name__ == '__main__':
    loss_output = LossOutput(**INIT_ARGS)
    result = getattr(loss_output, METHOD)(**ARGS)
    print(result)
    print(f"Float conversion: {float(loss_output)}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
MediumS3GCEncoder Ran
ran on a synthesized input
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("25ab1807c6cf8765")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, 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

# INIT_ARGS: parameters for MediumS3GCEncoder.__init__
INIT_ARGS = {
    'in_channels': 8,
    'hidden_channels': 6,
    'num_nodes': 10
}

# ARGS: parameters for MediumS3GCEncoder.forward
# AX and SX are precomputed graph convolutions with shape [num_nodes, in_channels]
# indices is optional for selecting a subset of nodes
ARGS = {
    'AX': torch.randn(10, 8),  # num_nodes=10, in_channels=8
    'SX': torch.randn(10, 8),  # num_nodes=10, in_channels=8
    'indices': torch.tensor([0, 2, 5, 7])  # optional subset of nodes
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
SmallS3GCEncoder Ran
ran on a synthesized input
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("d95c24e59cfbc319")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (3, 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
from torch import Tensor

class SmallS3GCEncoder(nn.Module):
    """
    Encoder for small-scale graphs (e.g., Cora).

    Uses simple linear transformations without hidden layers.

    Architecture:
        :math:`\\bar{X} = \\tilde{A}X\\Theta_1 + \\S_kX\\Theta_2 + I`
    """

    def __init__(self, in_channels: int, hidden_channels: int, num_nodes: int):
        super().__init__()
        self.w1 = nn.Linear(in_channels, hidden_channels, bias=True)
        self.w2 = nn.Linear(in_channels, hidden_channels, bias=True)
        self.iden = nn.Parameter(torch.randn(num_nodes, hidden_channels, dtype=torch.float))
        self.reset_parameters()

    def reset_parameters(self):
        self.w1.bias.data.fill_(0.0)
        self.w2.bias.data.fill_(0.0)
        nn.init.normal_(self.iden)

    def forward(self, AX: Tensor, SX: Tensor, indices=None) -> Tensor:
        if indices is not None:
            AX = AX[indices]
            SX = SX[indices]
            iden = self.iden[indices]
        else:
            iden = self.iden
        return torch.nn.functional.normalize(self.w1(AX) + self.w2(SX) + iden, p=2, dim=1)


INIT_ARGS = {
    "in_channels": 4,
    "hidden_channels": 3,
    "num_nodes": 5,
}

ARGS = {
    "AX": torch.randn(5, 4),
    "SX": torch.randn(5, 4),
    "indices": torch.tensor([0, 2, 4]),
}

if __name__ == "__main__":
    model = SmallS3GCEncoder(**INIT_ARGS)
    output = model(**ARGS)
    print("Output shape:", output.shape)
    print("Output:", output)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 (3, 3) — shape (3, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
BaseModel Not yet run
unverified (harvested, no recorded run)
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("185d768dac929853")
How it ran

no invocation recorded.

S3GC Not yet run
unverified (harvested, no recorded run)
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("4b5cdd4b2fdb7418")
How it ran

no invocation recorded.

TrainableModel Not yet run
unverified (harvested, no recorded run)
Cloudy1225/PyAGC/pyagc/models/s3gc.py
code served (permissive licence) · get_code("faa4eee371a76f7c")
How it ran

no invocation recorded.

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 2026-11-22.

Repositories linked to this paper

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

Abstract

Attributed Graph Clustering (AGC) is a fundamental unsupervised task that partitions nodes into cohesive groups by jointly modeling structural topology and node attributes. While the advent of graph neural networks and self-supervised learning has catalyzed a proliferation of AGC methodologies, a widening chasm persists between academic benchmark performance and the stringent demands of real-world industrial deployment. To bridge this gap, this survey provides a comprehensive, industrially grounded review of AGC from three complementary perspectives. First, we introduce the Encode-Cluster-Optimize taxonomic framework, which decomposes the diverse algorithmic landscape into three orthogonal, composable modules: representation encoding, cluster projection, and optimization strategy. This unified paradigm enables principled architectural comparisons and inspires novel methodological combinations. Second, we critically examine prevailing evaluation protocols to expose the field's academic monoculture: a pervasive overreliance on small, homophilous citation networks, the inadequacy of supervised-only metrics for an inherently unsupervised task, and the chronic neglect of computational scalability. In response, we advocate for a holistic evaluation standard that integrates supervised semantic alignment, unsupervised structural integrity, and rigorous efficiency profiling. Third, we explicitly confront the practical realities of industrial deployment. By analyzing operational constraints such as massive scale, severe heterophily, and tabular feature noise alongside extensive empirical evidence from our companion benchmark, we outline actionable engineering strategies. Furthermore, we chart a clear roadmap for future research, prioritizing heterophily-robust encoders, scalable joint optimization, and unsupervised model selection criteria to meet production-grade requirements. To support the community and ensure reproducibility, our industrial-scale benchmark library and a continuously updated reading list are openly available at https://github.com/Cloudy1225/PyAGC.

For agents

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

get_harvested_code_for_paper("2603.20829")
get_code_for_paper("2603.20829")
have("2603.20829")

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

curl https://syntology.ai/api/ran/2603.20829.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.20829.svg)](https://syntology.ai/paper/2603.20829)

Connect an agent — have() is free.