SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2309.16014 · 2023

Graph-level Representation Learning with Joint-Embedding Predictive Architectures

arXiv · PDF · Open in the Atlas

Code that ran

10 of 11 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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2309.16014.json · badge: /api/ran/2309.16014.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
geriskenderi/graph-jepa canonical (classifier label)
classifier basis: census
10 ran · 1 no recorded run
FunctionStatusWhere it lives
DiscreteEncoder Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/model_utils/feature_encoder.py
code served (permissive licence) · get_code("7f955ec40afcef19")
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 Embedding. 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)

# DiscreteEncoder takes two parameters:
# - nin: number of input categories (vocabulary size)
# - nhid: hidden dimension (embedding dimension)

ARGS = {
    "nin": 10,    # vocabulary size
    "nhid": 4,    # embedding dimension
}
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 Embedding — a Embedding output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Embedding 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
LapPE Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/data_utils/pe.py
pointer only (licence: MIT); the stored copy get_code serves, from LUMIA-Group/Cluster-wise-Graph-Transformer, is pointer only (licence: NONE) · get_code("2be5d5ffa7c2b792")
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 LUMIA-Group/Cluster-wise-Graph-Transformer, 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.

RWSE Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/data_utils/pe.py
pointer only (licence: MIT); the stored copy get_code serves, from LUMIA-Group/Cluster-wise-Graph-Transformer, is pointer only (licence: NONE) · get_code("e63f8bb61e55a99d")
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 LUMIA-Group/Cluster-wise-Graph-Transformer, 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.

cal_coarsen_adj Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/transform.py
pointer only (licence: MIT); the stored copy get_code serves, from LUMIA-Group/Cluster-wise-Graph-Transformer, is pointer only (licence: NONE) · get_code("586551d992664067")
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 LUMIA-Group/Cluster-wise-Graph-Transformer, 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.

combine_subgraphs Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/transform.py
code served (permissive licence) · get_code("6466f9f70510811c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Tensor, ndim 2, dtype family int, shape (2, 6) 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)

# Define graph structure
# Let's say we have a graph with 5 nodes and 8 edges
num_nodes = 5
num_edges = 8
num_selected = 3  # number of subgraphs

# edge_index: [2, num_edges] - standard graph edge representation
edge_index = torch.tensor([
    [0, 1, 1, 2, 2, 3, 3, 4],  # source nodes
    [1, 0, 2, 1, 3, 2, 4, 3]   # target nodes
], dtype=torch.long)

# subgraphs_nodes: [2, num_subgraph_nodes] 
# First row: which subgraph each node belongs to
# Second row: the actual node ID in the original graph
subgraphs_nodes = torch.tensor([
    [0, 0, 0, 1, 1, 1, 2, 2, 2],  # subgraph indices
    [0, 1, 2, 1, 2, 3, 2, 3, 4]   # node IDs
], dtype=torch.long)

# subgraphs_edges: [2, num_subgraph_edges]
# First row: which subgraph each edge belongs to
# Second row: the edge index in the original edge_index
subgraphs_edges = torch.tensor([
    [0, 0, 1, 1, 2, 2],  # subgraph indices
    [0, 2, 2, 4, 4, 6]   # edge indices
], dtype=torch.long)

ARGS = {
    'edge_index': edge_index,
    'subgraphs_nodes': subgraphs_nodes,
    'subgraphs_edges': subgraphs_edges,
    'num_selected': num_selected,
    'num_nodes': num_nodes
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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, 6) — shape (2, 6) 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
count_parameters Ran
ran, honoured the contract we drafted
geriskenderi/graph-jepa/core/trainer.py
code served (permissive licence) · get_code("f6b944f50d3f15ae")
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 int, ndim 0, dtype family int, 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)

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

model = SimpleModel()

ARGS = {
    "model": model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, 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
k_fold Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/trainer.py
code served (permissive licence) · get_code("3f97bfae7118b1d2")
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.data import Data, InMemoryDataset

torch.manual_seed(0)

class SimpleDataset(InMemoryDataset):
    def __init__(self):
        super().__init__(root=None)
        # Create a simple dataset with labels for stratification
        x = torch.randn(10, 4)  # 10 samples, 4 features
        y = torch.tensor([0, 1, 0, 1, 0, 1, 0, 1, 0, 1], dtype=torch.long)
        self.data = Data(x=x, y=y)
    
    def len(self):
        return 10
    
    def get(self, idx):
        return self.data

dataset = SimpleDataset()

ARGS = {
    "dataset": dataset,
    "folds": 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
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 — 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
random_walk Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/data_utils/pe.py
pointer only (licence: MIT); the stored copy get_code serves, from LUMIA-Group/Cluster-wise-Graph-Transformer, is pointer only (licence: NONE) · get_code("2d2ee0ffa615ecaf")
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 LUMIA-Group/Cluster-wise-Graph-Transformer, 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.

to_sparse Ran
ran, honoured the contract we drafted
geriskenderi/graph-jepa/core/transform.py
code served (permissive licence) · get_code("8039d4f24f678f0b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned tuple, ndim 3, dtype family int; 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

torch.manual_seed(0)

# node_mask: a boolean or binary tensor indicating which nodes are selected
# edge_mask: a boolean or binary tensor indicating which edges are selected
# The function calls .nonzero().T on both, so they should be tensors

# Create a small node mask (e.g., for 5 nodes across 2 subgraphs or similar structure)
node_mask = torch.tensor([
    [1, 0, 1, 0, 1],
    [0, 1, 1, 0, 0]
], dtype=torch.bool)

# Create a small edge mask (e.g., for 6 edges across 2 subgraphs or similar structure)
edge_mask = torch.tensor([
    [1, 0, 0, 1, 0, 1],
    [0, 1, 1, 0, 0, 0]
], dtype=torch.bool)

ARGS = {
    'node_mask': node_mask,
    'edge_mask': edge_mask
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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 3, dtype family int — 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
update_cfg Ran
ran; our contract draft was wrong, not the code
geriskenderi/graph-jepa/core/config.py
code served (permissive licence) · get_code("3764b11a07a3ebd8")
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 CfgNode. 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 argparse
from yacs.config import CfgNode as CN

def set_cfg(cfg):
    cfg.dataset = 'ZINC'
    cfg.num_workers = 8
    cfg.device = 0
    cfg.seed = None
    cfg.logfile = None
    cfg.depth = -1
    cfg.train = CN()
    cfg.train.batch_size = 128
    cfg.train.epochs = 1000
    cfg.train.runs = 4
    cfg.train.lr = 0.001
    cfg.train.lr_patience = 20
    cfg.train.lr_decay = 0.5
    cfg.train.wd = 0.0
    cfg.train.dropout = 0.0
    cfg.train.mlpmixer_dropout = 0.0
    cfg.train.min_lr = 1e-05
    cfg.train.optimizer = 'Adam'
    cfg.train.multiscale = False
    cfg.model = CN()
    cfg.model.gnn_type = 'GINEConv'
    cfg.model.gMHA_type = 'MLPMixer'
    cfg.model.hidden_size = 128
    cfg.model.nlayer_gnn = 4
    cfg.model.nlayer_mlpmixer = 4
    cfg.model.pool = 'mean'
    cfg.model.residual = True
    cfg.pos_enc = CN()
    cfg.pos_enc.rw_dim = 0
    cfg.pos_enc.lap_dim = 0
    cfg.pos_enc.patch_rw_dim = 8
    cfg.pos_enc.patch_num_diff = -1
    cfg.metis = CN()
    cfg.metis.enable = True
    cfg.metis.online = True
    cfg.metis.n_patches = 32
    cfg.metis.drop_rate = 0.3
    cfg.metis.num_hops = 1
    cfg.jepa = CN()
    cfg.jepa.enable = True
    cfg.jepa.num_context = 1
    cfg.jepa.num_targets = 4
    cfg.jepa.dist = 0
    return cfg

def update_cfg(cfg, args_str=None):
    parser = argparse.ArgumentParser()
    parser.add_argument('--config', default='', metavar='FILE', help='Path to config file')
    parser.add_argument('opts', default=[], nargs=argparse.REMAINDER, help='Modify config options using the command-line')
    if isinstance(args_str, str):
        args = parser.parse_args(args_str.split())
    else:
        args = parser.parse_args()
    cfg = cfg.clone()
    if os.path.isfile(args.config):
        cfg.merge_from_file(args.config)
    cfg.merge_from_list(args.opts)
    return cfg

cfg = set_cfg(CN())

ARGS = {
    'cfg': cfg,
    'args_str': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yacs — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type CfgNode — a CfgNode output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a CfgNode 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
config_logger Not yet run
unverified (harvested, no recorded run)
geriskenderi/graph-jepa/core/log.py
code served (permissive licence) · get_code("7a15a3fa5bfa6eab")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 10 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 10 later) and 1 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
count_parameters 180 other papers
100 dated earlier, 76 later, 4 same year, order not determinable · 8 only as an assembly brick
1705.08415 (2017-05)
Supervised Community Detection with Line Graph Neural Networks
1707.01926 (2017-07)
Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting
1808.09367 (2018-08)
Deriving Machine Attention from Human Rationales
and 177 more not named here
update_cfg 8 other papers
3 dated earlier, 5 later
2110.03753 (2021-10)
From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness
2304.07702 (2023-04)
An Empirical Study of Realized GNN Expressiveness
2305.19523 (2023-05)
Harnessing Explanations: LLM-to-LM Interpreter for Enhanced Text-Attributed Graph Representation Learning
and 5 more not named here
combine_subgraphs 3 other papers
1 dated earlier, 2 later
2210.09521 (2022-10)
A Practical, Progressively-Expressive GNN
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
to_sparse 3 other papers
1 dated earlier, 2 later
2210.09521 (2022-10)
A Practical, Progressively-Expressive GNN
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
LapPE 2 other papers
2 later
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
RWSE 2 other papers
2 later
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
cal_coarsen_adj 2 other papers
2 later
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
k_fold 2 other papers
1 dated earlier, 1 later
2110.03753 (2021-10)
From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
random_walk 2 other papers
2 later
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in Transformers!
2410.06746 (2024-10)
Cluster-wise Graph Transformer with Dual-granularity Kernelized Attention
config_logger 1 other paper
1 later
2312.04234 (2023-12)
Graph Convolutions Enrich the Self-Attention in 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

Joint-Embedding Predictive Architectures (JEPAs) have recently emerged as a novel and powerful technique for self-supervised representation learning. They aim to learn an energy-based model by predicting the latent representation of a target signal y from the latent representation of a context signal x. JEPAs bypass the need for negative and positive samples, traditionally required by contrastive learning while avoiding the overfitting issues associated with generative pretraining. In this paper, we show that graph-level representations can be effectively modeled using this paradigm by proposing a Graph Joint-Embedding Predictive Architecture (Graph-JEPA). In particular, we employ masked modeling and focus on predicting the latent representations of masked subgraphs starting from the latent representation of a context subgraph. To endow the representations with the implicit hierarchy that is often present in graph-level concepts, we devise an alternative prediction objective that consists of predicting the coordinates of the encoded subgraphs on the unit hyperbola in the 2D plane. Through multiple experimental evaluations, we show that Graph-JEPA can learn highly semantic and expressive representations, as shown by the downstream performance in graph classification, regression, and distinguishing non-isomorphic graphs. The code is available at https://github.com/geriskenderi/graph-jepa.

For agents

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

get_harvested_code_for_paper("2309.16014")
get_code_for_paper("2309.16014")
have("2309.16014")

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

curl https://syntology.ai/api/ran/2309.16014.json

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

[![Syntology run record](https://syntology.ai/api/ran/2309.16014.svg)](https://syntology.ai/paper/2309.16014)

Connect an agent — have() is free.