SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2010.02863 · 2020

Directional Graph Networks

arXiv · PDF · Open in the Atlas

Code that ran

13 of 14 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; 3 honoured the contract we drafted, 1 violated it and 7 have no contract verdict; 2 failures of our instrument; 1 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/2010.02863.json · badge: /api/ran/2010.02863.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
gbouritsas/gsn community (classifier label)
classifier basis not recorded on the link
13 ran · 1 no recorded run
FunctionStatusWhere it lives
aggregate_max Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/nets/aggregators.py
code served (permissive licence) · get_code("ee86c81b47055403")
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 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

torch.manual_seed(0)

# The function uses:
# - h: needs dim=1 to be indexable, so shape (batch, features, ...)
# - vector_field: not used in the function body
# - h_in: not used in the function body

# Create small tensors
h = torch.randn(2, 4)  # batch_size=2, features=4
vector_field = torch.randn(2, 3)  # arbitrary, not used
h_in = torch.randn(2, 5)  # arbitrary, not used

ARGS = {
    "h": h,
    "vector_field": vector_field,
    "h_in": h_in
}
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 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
aggregate_mean Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/nets/aggregators.py
code served (permissive licence) · get_code("24fa352e05630952")
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 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

torch.manual_seed(0)

# The function uses:
# - h: torch.mean() is called on it with dim=1, so it needs to be a 2D tensor
# - vector_field: not used in the function body
# - h_in: not used in the function body

h = torch.randn(2, 4)
vector_field = torch.randn(2, 3)
h_in = torch.randn(2, 5)

ARGS = {
    "h": h,
    "vector_field": vector_field,
    "h_in": h_in,
}
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 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
aggregate_min Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/nets/aggregators.py
code served (permissive licence) · get_code("a79bba4bc6411d00")
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 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Analyze the function:
# - h: tensor passed to torch.min(h, dim=1)[0]
# - vector_field: not used in function body
# - h_in: not used in function body
# 
# torch.min(h, dim=1)[0] requires h to be at least 2D
# The function takes min along dim=1, so h needs shape (batch_size, feature_dim, ...)

# Create a small 3D tensor: (batch=2, dim=4, seq=3)
h = torch.randn(2, 4, 3)

# vector_field is not used, but we need to provide something
# Make it a small tensor
vector_field = torch.randn(2, 4)

# h_in is not used, but we need to provide something
# Make it a small tensor
h_in = torch.randn(2, 4)

ARGS = {
    "h": h,
    "vector_field": vector_field,
    "h_in": h_in,
}
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 2, dtype family float, shape (2, 3) — shape (2, 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
get_activation Ran
ran; our contract draft was wrong, not the code
gbouritsas/gsn/directional_gsn/nets/layers.py
pointer only (licence: MIT); the stored copy get_code serves, from Saro00/DGN, is pointer only (licence: NONE) · get_code("fe76e55f0b804b7b")
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 Saro00/DGN, 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.

get_custom_edge_list Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/utils_subgraph_encoding.py
code served (permissive licence) · get_code("8c86e6abd3cdd4c4")
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 list. 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
import types

# Test with substructure_type parameter
# Using a NetworkX graph generator that works with an integer parameter
ARGS = {
    'ks': [3, 4],
    'substructure_type': 'complete_graph',
    'filename': None
}
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 list — a list 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
gpu_setup Ran
ran; our contract draft was wrong, not the code
gbouritsas/gsn/directional_gsn/main_HIV.py
pointer only (licence: MIT); the stored copy get_code serves, from trustworthygnn/mia-gnn, is pointer only (licence: NONE) · get_code("f72434c0dcd5d11b")
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 trustworthygnn/mia-gnn, 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.

scale_amplification Ran
ran, honoured the contract we drafted
gbouritsas/gsn/directional_gsn/nets/scalers.py
pointer only (licence: MIT); the stored copy get_code serves, from changminwu/expandergnn, is pointer only (licence: NONE) · get_code("5d07dbe669b9f73a")
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 changminwu/expandergnn, 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.

scale_attenuation Ran
ran, honoured the contract we drafted
gbouritsas/gsn/directional_gsn/nets/scalers.py
pointer only (licence: MIT); the stored copy get_code serves, from changminwu/expandergnn, is pointer only (licence: NONE) · get_code("d0d453f04b18e8f1")
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 changminwu/expandergnn, 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.

scale_identity Ran
ran, honoured the contract we drafted
gbouritsas/gsn/directional_gsn/nets/scalers.py
pointer only (licence: MIT); the stored copy get_code serves, from changminwu/expandergnn, is pointer only (licence: NONE) · get_code("2d32c33e16bad56e")
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 changminwu/expandergnn, 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.

str2ListOfLists2int Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/utils_parsing.py
code served (permissive licence) · get_code("94c293b9defb4188")
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 list. 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.

# Analyze the function to understand what it expects
# str2ListOfLists2int(v) takes a string parameter v
# It splits by ',,', then splits each part by ',', then converts each char to int (or [] if space)

# Let's create a valid input string
v = "1,2,3,,4, ,5,,6,7"

ARGS = {
    "v": v
}
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 list — a list 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
str2ListOfListsOfLists2int Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/utils_parsing.py
code served (permissive licence) · get_code("fa9cf77a40a3bfae")
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 list. 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

def str2ListOfListsOfLists2int(v):
    return [[[[] if c == ' ' else int(c) for c in vii.split(',')] for vii in vi.split(',,')] for vi in v.split(',,,')]

# The function takes a string parameter 'v' and parses it hierarchically
# Split by ',,,' for outer list
# Split by ',,' for middle list
# Split by ',' for inner list
# Convert non-space characters to int

v_test = "1,2,3,,4,5,,6 ,7,,,8,9,,1,,,2,3"

ARGS = {
    'v': v_test
}
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 list — a list 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
str2bool Ran
ran, violated the contract we drafted
gbouritsas/gsn/directional_gsn/utils_parsing.py
code served (permissive licence) · get_code("f017532fc389cbfe")
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) and returned bool, 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 argparse

def str2bool(v):
    if v.lower() in ('yes', 'true', 't', 'y', '1'):
        return True
    elif v.lower() in ('no', 'false', 'f', 'n', '0'):
        return False
    else:
        raise argparse.ArgumentTypeError('Boolean value expected.')

# Setup code to construct valid arguments
ARGS = {
    'v': 'true'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 bool, 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
to_undirected Ran
ran on a synthesized input
gbouritsas/gsn/directional_gsn/utils_graph_processing.py
code served (permissive licence) · get_code("77a61130d1c9fd6b")
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 2, dtype family int, shape (2, 4) 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

torch.manual_seed(0)

# edge_index is a tuple of two tensors representing source and target node indices
# For a small graph, we'll create edge indices with 2 edges
row = torch.tensor([0, 1], dtype=torch.long)
col = torch.tensor([1, 2], dtype=torch.long)
edge_index = (row, col)

ARGS = {
    "edge_index": edge_index
}
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 2, dtype family int, shape (2, 4) — shape (2, 4) 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
encode Not yet run
unverified (harvested, no recorded run)
gbouritsas/gsn/directional_gsn/utils_one_hot_encoding.py
code served (permissive licence) · get_code("831368bc983a71e0")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 7 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 0 the same month, 3 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.

FunctionAlso present inEarliest-datedOthers named
str2bool 82 other papers
39 dated earlier, 43 later
1510.03055 (2015-10)
A Diversity-Promoting Objective Function for Neural Conversation Models
1703.06490 (2017-03)
Generating Multi-label Discrete Patient Records using Generative Adversarial Networks
1710.10467 (2017-10)
Generalized End-to-End Loss for Speaker Verification
and 79 more not named here
gpu_setup 8 other papers
2 dated earlier, 6 later
2003.00982 (2020-03)
Benchmarking Graph Neural Networks
2009.11746 (2020-09)
Learning Graph Normalization for Graph Neural Networks
2012.10258 (2020-12)
An Experimental Study of the Transferability of Spectral Graph Networks
and 5 more not named here
get_activation 2 other papers
1 dated earlier, 1 later
2004.05718 (2020-04)
Principal Neighbourhood Aggregation for Graph Nets
2409.19414 (2024-09)
Sequential Signal Mixing Aggregation for Message Passing Graph Neural Networks
scale_amplification 1 other paper
1 dated earlier
2004.05718 (2020-04)
Principal Neighbourhood Aggregation for Graph Nets
scale_attenuation 1 other paper
1 dated earlier
2004.05718 (2020-04)
Principal Neighbourhood Aggregation for Graph Nets
scale_identity 1 other paper
1 dated earlier
2004.05718 (2020-04)
Principal Neighbourhood Aggregation for Graph Nets
to_undirected 1 other paper
1 dated earlier
2006.09252 (2020-06)
Improving Graph Neural Network Expressivity via Subgraph Isomorphism Counting

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

The lack of anisotropic kernels in graph neural networks (GNNs) strongly limits their expressiveness, contributing to well-known issues such as over-smoothing. To overcome this limitation, we propose the first globally consistent anisotropic kernels for GNNs, allowing for graph convolutions that are defined according to topologicaly-derived directional flows. First, by defining a vector field in the graph, we develop a method of applying directional derivatives and smoothing by projecting node-specific messages into the field. Then, we propose the use of the Laplacian eigenvectors as such vector field. We show that the method generalizes CNNs on an $n$-dimensional grid and is provably more discriminative than standard GNNs regarding the Weisfeiler-Lehman 1-WL test. We evaluate our method on different standard benchmarks and see a relative error reduction of 8% on the CIFAR10 graph dataset and 11% to 32% on the molecular ZINC dataset, and a relative increase in precision of 1.6% on the MolPCBA dataset. An important outcome of this work is that it enables graph networks to embed directions in an unsupervised way, thus allowing a better representation of the anisotropic features in different physical or biological problems.

For agents

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

get_harvested_code_for_paper("2010.02863")
get_code_for_paper("2010.02863")
have("2010.02863")

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

curl https://syntology.ai/api/ran/2010.02863.json

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

[![Syntology run record](https://syntology.ai/api/ran/2010.02863.svg)](https://syntology.ai/paper/2010.02863)

Connect an agent — have() is free.