SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · aaai_26026 · AAAI · 2023

Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks

Kun He, Chao Li, Hao Xu

Open in the Atlas

Code that ran

5 of 5 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 2 have no contract verdict; 3 failures of our instrument; 0 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/aaai_26026.json · badge: /api/ran/aaai_26026.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
JHL-HUST/PMMM canonical (classifier label)
classifier basis: census
5 ran · 0 no recorded run
FunctionStatusWhere it lives
normalize_row Ran
ran; our contract draft was wrong, not the code
JHL-HUST/PMMM/node_classification/preprocess.py
code served (permissive licence) · get_code("fbdd0e21bd4c2b07")
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 coo_matrix. 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
import scipy.sparse as sp

# Create a small sparse matrix for testing
# The function expects a sparse matrix with row-wise operations
# Create a 4x3 sparse matrix with some non-zero values
data = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
row = np.array([0, 0, 1, 2, 3])
col = np.array([0, 1, 2, 1, 0])
mx = sp.csr_matrix((data, (row, col)), shape=(4, 3))

ARGS = {
    "mx": mx
}
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 coo_matrix — a coo_matrix output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coo_matrix 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
normalize_sym Ran
ran; our contract draft was wrong, not the code
JHL-HUST/PMMM/node_classification/preprocess.py
code served (permissive licence) · get_code("04bb317b2ecbb18a")
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 coo_matrix. 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
import scipy.sparse as sp

# Create a small sparse adjacency matrix
# Using a 4x4 sparse matrix to test the normalize_sym function
np.random.seed(0)
adj_dense = np.random.rand(4, 4)
adj_dense = (adj_dense > 0.5).astype(float)  # Make it sparse-like
adj_sparse = sp.csr_matrix(adj_dense)

ARGS = {
    "adj": adj_sparse
}
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 coo_matrix — a coo_matrix output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coo_matrix 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
sparse_mx_to_torch_sparse_tensor Ran
ran; our contract draft was wrong, not the code
JHL-HUST/PMMM/node_classification/preprocess.py
code served (permissive licence) · get_code("9378be67198d5141")
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. 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
import torch
from scipy import sparse

torch.manual_seed(0)

# Create a small scipy sparse matrix in COO format
# The function expects sparse_mx to have .row, .col, .data, and .shape attributes
row = np.array([0, 1, 2])
col = np.array([1, 2, 0])
data = np.array([1.5, 2.5, 3.5])
sparse_mx = sparse.coo_matrix((data, (row, col)), shape=(3, 3))

ARGS = {
    "sparse_mx": sparse_mx
}
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, shape (3, 3) — shape (3, 3) as a later run recorded it
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 Ran
ran on a synthesized input
JHL-HUST/PMMM/node_classification/train_search.py
code served (permissive licence) · get_code("d8723a8b0ad4ba47")
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
import torch.optim as optim

torch.manual_seed(0)

# Define a minimal model that satisfies the function's usage
class SimpleModel(nn.Module):
    def __init__(self, num_node_types, num_classes):
        super().__init__()
        self.linear = nn.Linear(4 + num_node_types, num_classes)
    
    def forward(self, node_feats, node_types, adjs):
        # node_feats: (num_nodes, 4)
        # node_types: (num_nodes, num_node_types)
        # adjs: list of adjacency matrices
        # Combine features and types, pass through linear layer
        combined = torch.cat([node_feats, node_types], dim=1)
        out = self.linear(combined)
        return out

# Create small tensors
num_nodes = 2
num_features = 4
num_node_types = 3
num_classes = 2

node_feats = torch.randn(num_nodes, num_features)
node_types = torch.randn(num_nodes, num_node_types)

# Create a simple adjacency matrix
adjs = [torch.randn(num_nodes, num_nodes)]

# Create indices and targets
train_idx = torch.tensor([0], dtype=torch.long)
train_target = torch.tensor([0], dtype=torch.long)

valid_idx = torch.tensor([1], dtype=torch.long)
valid_target = torch.tensor([1], dtype=torch.long)

# Create model and optimizers
model = SimpleModel(num_node_types, num_classes)
optimizer_w = optim.SGD(model.parameters(), lr=0.01)
optimizer_a = optim.SGD(model.parameters(), lr=0.01)

eps = 1e-5

ARGS = {
    'node_feats': node_feats,
    'node_types': node_types,
    'adjs': adjs,
    'train_idx': train_idx,
    'train_target': train_target,
    'valid_idx': valid_idx,
    'valid_target': valid_target,
    'model': model,
    'optimizer_w': optimizer_w,
    'optimizer_a': optimizer_a,
    'eps': eps
}
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, 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
train Ran
ran on a synthesized input
JHL-HUST/PMMM/recommendation/train_search.py
code served (permissive licence) · get_code("b5bc094ba8fd1b6b")
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
import torch.optim as optim

torch.manual_seed(0)

# Number of nodes and embedding dimension
num_nodes = 8
embedding_dim = 4
num_node_types = 3
num_edge_types = 2

# Create node features (num_nodes x feature_dim)
node_feats = torch.randn(num_nodes, 5)

# Create node types (num_nodes,)
node_types = torch.randint(0, num_node_types, (num_nodes,))

# Create adjacency matrices (list of sparse representations or dense)
# Simple: list of adjacency matrices, one per edge type
adjs = [torch.randn(num_nodes, num_nodes) for _ in range(num_edge_types)]

# Create positive and negative train/val edge indices
# Shape: (batch_size, 2) where each row is [source_node_idx, target_node_idx]
pos_train = torch.tensor([[0, 1], [2, 3]], dtype=torch.long)
neg_train = torch.tensor([[1, 0], [3, 2]], dtype=torch.long)

pos_val = torch.tensor([[4, 5], [6, 7]], dtype=torch.long)
neg_val = torch.tensor([[5, 4], [7, 6]], dtype=torch.long)

# Define a minimal GNN model that takes (node_feats, node_types, adjs) and outputs embeddings
class SimpleGNNModel(nn.Module):
    def __init__(self, input_dim, output_dim):
        super().__init__()
        self.linear = nn.Linear(input_dim, output_dim)
    
    def forward(self, node_feats, node_types, adjs):
        # Simple forward: just apply linear transformation to node features
        # Output shape: (num_nodes, output_dim)
        return self.linear(node_feats)

# Instantiate models
model_s = SimpleGNNModel(input_dim=5, output_dim=embedding_dim)
model_t = SimpleGNNModel(input_dim=5, output_dim=embedding_dim)

# Create optimizers for model parameters
optimizer_w = optim.SGD(list(model_s.parameters()) + list(model_t.parameters()), lr=0.01)
optimizer_a = optim.SGD(list(model_s.parameters()) + list(model_t.parameters()), lr=0.01)

# Epsilon parameter
eps = 1e-8

ARGS = {
    "node_feats": node_feats,
    "node_types": node_types,
    "adjs": adjs,
    "pos_train": pos_train,
    "neg_train": neg_train,
    "pos_val": pos_val,
    "neg_val": neg_val,
    "model_s": model_s,
    "model_t": model_t,
    "optimizer_w": optimizer_w,
    "optimizer_a": optimizer_a,
    "eps": eps,
}
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, 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

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 distinct functions with an identity, 3 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 3 later) and 2 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
normalize_row 1 other paper
1 later
2402.11518 (2024-02)
Large Language Model-driven Meta-structure Discovery in Heterogeneous Information Network
normalize_sym 1 other paper
1 later
2402.11518 (2024-02)
Large Language Model-driven Meta-structure Discovery in Heterogeneous Information Network
sparse_mx_to_torch_sparse_tensor 1 other paper
1 later
2402.11518 (2024-02)
Large Language Model-driven Meta-structure Discovery in Heterogeneous Information Network

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

Heterogeneous information networks (HINs) are widely employed for describing real-world data with intricate entities and relationships. To automatically utilize their semantic information, graph neural architecture search has recently been developed on various tasks of HINs. Existing works, on the other hand, show weaknesses in instability and inflexibility. To address these issues, we propose a novel method called Partial Message Meta Multigraph search (PMMM) to automatically optimize the neural architecture design on HINs. Specifically, to learn how graph neural networks (GNNs) propagate messages along various types of edges, PMMM adopts an efficient differentiable framework to search for a meaningful meta multigraph, which can capture more flexible and complex semantic relations than a meta graph. The differentiable search typically suffers from performance instability, so we further propose a stable algorithm called partial message search to ensure that the searched meta multigraph consistently surpasses the manually designed meta-structures, i.e., meta-paths. Extensive experiments on six benchmark datasets over two representative tasks, including node classification and recommendation, demonstrate the effectiveness of the proposed method. Our approach outperforms the stateof-the-art heterogeneous GNNs, finds out meaningful meta multigraphs, and is significantly more stable. Our code is available at https://github.com/JHL-HUST/PMMM.

For agents

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

get_harvested_code_for_paper("aaai_26026")
get_code_for_paper("aaai_26026")
have("aaai_26026")

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

curl https://syntology.ai/api/ran/aaai_26026.json

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

[![Syntology run record](https://syntology.ai/api/ran/aaai_26026.svg)](https://syntology.ai/paper/aaai_26026)

Connect an agent — have() is free.