SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.00596 · 2026

Kernelized Edge Attention: Addressing Semantic Attention Blurring in Temporal Graph Neural Networks

Srikanta Bedathur, Sonia Gupta, Govind Waghmare, Srini Rohan, Gujulla Leel, Nikhil Tumbde, Mastercard

arXiv · PDF · Open in the Atlas

Code that ran

14 of 21 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (6 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 4 failures of our instrument; 7 no recorded run), measured 2026-11-21; 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/2602.00596.json · badge: /api/ran/2602.00596.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
storyandwine/GEARSage-DGraphFin canonical (classifier label)
classifier basis: census
5 ran · 4 no recorded run
shenyangHuang/TGB community
Papers with Code archive
8 ran · 3 no recorded run
yule-BUAA/DyGLib community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
EdgePredictor_per_node Ran
ran on a synthesized input
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("3a203beb102a8b30")
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 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
import numpy as np

# INIT_ARGS for EdgePredictor_per_node.__init__(dim_in_time, dim_in_node, predict_class)
INIT_ARGS = {
    'dim_in_time': 3,
    'dim_in_node': 5,
    'predict_class': 2
}

# ARGS for EdgePredictor_per_node.forward(h, neg_samples=1)
# h.shape[0] must be divisible by (neg_samples + 2)
# h.shape[1] must equal dim_in_time + dim_in_node = 3 + 5 = 8
neg_samples = 2
num_edge = 4
total_nodes = num_edge * (neg_samples + 2)  # 4 * 4 = 16

h = torch.randn(total_nodes, INIT_ARGS['dim_in_time'] + INIT_ARGS['dim_in_node'])

ARGS = {
    'h': h,
    'neg_samples': neg_samples
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy, torch — 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
FeatEncode Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("39c0adbee26233ea")
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 an instance of its own class, FeatEncode: the run constructed the object and did not call it, so no result was computed. 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
import numpy as np

torch.manual_seed(0)

# Create small test tensors
batch_size = 2
time_dims = 4
feat_dims = 3
out_dims = 5

# edge_feats: batch_size x feat_dims
edge_feats = torch.randn(batch_size, feat_dims)

# edge_ts: batch_size (timestamps)
edge_ts = torch.randn(batch_size)

ARGS = {
    "time_dims": time_dims,
    "feat_dims": feat_dims,
    "out_dims": out_dims,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type FeatEncode — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a FeatEncode 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
FeedForward Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("f512e4efa2c5957a")
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 an instance of its own class, FeedForward: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn.functional as F

torch.manual_seed(0)

class FeedForward(nn.Module):
    """
    2-layer MLP with GeLU (fancy version of ReLU) as activation
    """

    def __init__(self, dims, expansion_factor, dropout=0, use_single_layer=False):
        super().__init__()
        self.dims = dims
        self.use_single_layer = use_single_layer
        self.expansion_factor = expansion_factor
        self.dropout = dropout
        if use_single_layer:
            self.linear_0 = nn.Linear(dims, dims)
        else:
            self.linear_0 = nn.Linear(dims, int(expansion_factor * dims))
            self.linear_1 = nn.Linear(int(expansion_factor * dims), dims)
        self.reset_parameters()

    def reset_parameters(self):
        self.linear_0.reset_parameters()
        if self.use_single_layer == False:
            self.linear_1.reset_parameters()

    def forward(self, x):
        x = self.linear_0(x)
        x = F.gelu(x)
        x = F.dropout(x, p=self.dropout, training=self.training)
        if self.use_single_layer == False:
            x = self.linear_1(x)
            x = F.dropout(x, p=self.dropout, training=self.training)
        return x


ARGS = {
    "dims": 4,
    "expansion_factor": 2,
    "dropout": 0.1,
    "use_single_layer": False
}
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 FeedForward — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a FeedForward 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
MultiHeadAttention Ran
ran on a synthesized input
yule-BUAA/DyGLib/models/modules.py
code served (permissive licence) · get_code("d6264ffe9f545d50")
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 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
import numpy as np

INIT_ARGS = {
    'node_feat_dim': 4,
    'edge_feat_dim': 3,
    'time_feat_dim': 2,
    'num_heads': 2,
    'dropout': 0.1
}

# Create sample tensors and masks
batch_size = 2
num_neighbors = 5

node_features = torch.randn(batch_size, INIT_ARGS['node_feat_dim'])
node_time_features = torch.randn(batch_size, 1, INIT_ARGS['time_feat_dim'])
neighbor_node_features = torch.randn(batch_size, num_neighbors, INIT_ARGS['node_feat_dim'])
neighbor_node_time_features = torch.randn(batch_size, num_neighbors, INIT_ARGS['time_feat_dim'])
neighbor_node_edge_features = torch.randn(batch_size, num_neighbors, INIT_ARGS['edge_feat_dim'])
neighbor_masks = np.ones((batch_size, num_neighbors))
neighbor_masks[:, -1] = 0  # Mask out last neighbor for each batch

ARGS = {
    'node_features': node_features,
    'node_time_features': node_time_features,
    'neighbor_node_features': neighbor_node_features,
    'neighbor_node_time_features': neighbor_node_time_features,
    'neighbor_node_edge_features': neighbor_node_edge_features,
    'neighbor_masks': neighbor_masks
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy, torch — 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
PositionalEncoding1D Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("415730aff961e7cd")
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 an instance of its own class, PositionalEncoding1D: the run constructed the object and did not call it, so no result was computed. 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
import numpy as np

torch.manual_seed(0)

# PositionalEncoding1D takes only one parameter: channels
# Looking at the forward method, it expects a 3d tensor of shape (batch_size, x, ch)
# where ch should match or be related to the channels parameter

channels = 8

ARGS = {
    'channels': channels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PositionalEncoding1D — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PositionalEncoding1D 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
Summer Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("4976a15c63f01726")
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 an instance of its own class, Summer: the run constructed the object and did not call it, so no result was computed. 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)

# Create a minimal positional encoding module that:
# 1. Is callable (has __call__ method from nn.Module)
# 2. Takes a tensor as input
# 3. Returns a tensor of the same size as the input
class PositionalEncoding(nn.Module):
    def __init__(self):
        super(PositionalEncoding, self).__init__()
    
    def forward(self, tensor):
        # Return a tensor of zeros with the same shape as input
        # This satisfies the size matching requirement
        return torch.zeros_like(tensor)

# Instantiate the positional encoding module
penc = PositionalEncoding()

ARGS = {
    "penc": penc
}
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 Summer — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Summer 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
TimeEncode Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("ddd8a5727b202d18")
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 an instance of its own class, TimeEncode: the run constructed the object and did not call it, so no result was computed. 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
import numpy as np

torch.manual_seed(0)

# TimeEncode.__init__ takes one parameter: dim
# TimeEncode.forward takes one parameter: t (a tensor of timestamps)

ARGS = {
    "dim": 4,
}

# Create an instance to verify it works
model = nn.Module.__new__(type('TimeEncode', (nn.Module,), {
    '__init__': lambda self, dim: (
        nn.Module.__init__(self),
        setattr(self, 'dim', dim),
        setattr(self, 'w', nn.Linear(1, dim)),
        self.reset_parameters()
    ),
    'reset_parameters': lambda self: (
        setattr(self.w, 'weight', nn.Parameter(torch.from_numpy(1 / 10 ** np.linspace(0, 9, self.dim, dtype=np.float32)).reshape(self.dim, -1))),
        setattr(self.w, 'bias', nn.Parameter(torch.zeros(self.dim))),
        setattr(self.w.weight, 'requires_grad', False),
        setattr(self.w.bias, 'requires_grad', False)
    ),
    'forward': torch.no_grad()(lambda self, t: torch.cos(self.w(t.reshape((-1, 1)))))
}))

# Create test tensor for forward pass
test_t = torch.randn(2, 1)
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TimeEncode — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TimeEncode 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
_MultiheadAttention Ran
ran on a synthesized input · constructed an object; no result was computed
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("3817d890b2e0ce53")
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 an instance of its own class, _MultiheadAttention: the run constructed the object and did not call it, so no result was computed. 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
import numpy as np

torch.manual_seed(0)

# Parameters for _MultiheadAttention
d_model = 8
n_heads = 2
d_k = 4
d_v = 4
res_attention = False
attn_dropout = 0.1
proj_dropout = 0.1
qkv_bias = True
lsa = False

# Create sample tensors for forward pass testing
batch_size = 2
seq_len = 3
Q = torch.randn(batch_size, seq_len, d_model)
K = torch.randn(batch_size, seq_len, d_model)
V = torch.randn(batch_size, seq_len, d_model)
prev = None
key_padding_mask = None
attn_mask = None

ARGS = {
    'd_model': d_model,
    'n_heads': n_heads,
    'd_k': d_k,
    'd_v': d_v,
    'res_attention': res_attention,
    'attn_dropout': attn_dropout,
    'proj_dropout': proj_dropout,
    'qkv_bias': qkv_bias,
    'lsa': lsa,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type _MultiheadAttention — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a _MultiheadAttention 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
_ScaledDotProductAttention Ran
ran on a synthesized input
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("d253e560e9ac34a1")
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 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
import numpy as np

INIT_ARGS = {
    'd_model': 4,
    'n_heads': 2,
    'attn_dropout': 0.1,
    'res_attention': True,
    'lsa': True
}

# Create input tensors matching the expected shapes
bs = 2
seq_len = 3
d_k = INIT_ARGS['d_model'] // INIT_ARGS['n_heads']

q = torch.randn(bs, INIT_ARGS['n_heads'], seq_len, d_k)
k = torch.randn(bs, INIT_ARGS['n_heads'], d_k, seq_len)
v = torch.randn(bs, INIT_ARGS['n_heads'], seq_len, d_k)
prev = torch.randn(bs, INIT_ARGS['n_heads'], seq_len, seq_len)
key_padding_mask = torch.zeros(bs, seq_len, dtype=torch.bool)
key_padding_mask[:, -1] = True  # Mask last token for each batch
attn_mask = torch.triu(torch.ones(seq_len, seq_len) * -np.inf, diagonal=1)

ARGS = {
    'q': q,
    'k': k,
    'v': v,
    'prev': prev,
    'key_padding_mask': key_padding_mask,
    'attn_mask': attn_mask
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy, torch — 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
add_degree_feature Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
storyandwine/GEARSage-DGraphFin/feat_func.py
pointer only (licence: NONE) · get_code("0f8414b56c3a3051")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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.

add_feature_flag Ran
ran; our contract draft was wrong, not the code
storyandwine/GEARSage-DGraphFin/feat_func.py
pointer only (licence: NONE) · get_code("357d9970dfef8177")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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.

creat_activation_layer Ran
ran; our contract draft was wrong, not the code
storyandwine/GEARSage-DGraphFin/models/model.py
pointer only (licence: NONE) · get_code("743cb7598629fe21")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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.

degree_feat Ran
ran, honoured the contract we drafted
storyandwine/GEARSage-DGraphFin/utils/dgraphfin.py
pointer only (licence: NONE) · get_code("5445a93ec745cd56")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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.

edge_type_feat Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
storyandwine/GEARSage-DGraphFin/utils/dgraphfin.py
pointer only (licence: NONE) · get_code("40804320d6cd06ef")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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.

Patch_Encoding Not yet run
unverified (harvested, no recorded run)
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("850d001f29ffa49f")
How it ran

no invocation recorded.

STHN_Interface Not yet run
unverified (harvested, no recorded run)
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("64f196855787fc5c")
How it ran

no invocation recorded.

TransformerBlock Not yet run
unverified (harvested, no recorded run)
shenyangHuang/TGB/modules/sthn.py
code served (permissive licence) · get_code("c0c8de46adf7473e")
How it ran

no invocation recorded.

add_label_feature Not yet run
unverified (harvested, no recorded run)
storyandwine/GEARSage-DGraphFin/feat_func.py
pointer only (licence: NONE) · get_code("259ffd86dd711857")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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. No invocation recorded.

edge_timestamp_feat Not yet run
unverified (harvested, no recorded run)
storyandwine/GEARSage-DGraphFin/utils/dgraphfin.py
pointer only (licence: NONE) · get_code("b270f7d1c1270190")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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. No invocation recorded.

prepare_folder Not yet run
unverified (harvested, no recorded run)
storyandwine/GEARSage-DGraphFin/utils/utils.py
pointer only (licence: NONE) · get_code("df243d1bf9d82190")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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. No invocation recorded.

prepare_tune_folder Not yet run
unverified (harvested, no recorded run)
storyandwine/GEARSage-DGraphFin/utils/utils.py
pointer only (licence: NONE) · get_code("429b326513a25e02")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, storyandwine/GEARSage-DGraphFin 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. No invocation recorded.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 1 later) and 19 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
_ScaledDotProductAttention
brick of this paper's assembly: helper
4 other papers
2 dated earlier, 1 the same month, 1 later · 2 only as an assembly brick
2211.14730 (2022-11)
A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
2502.16284 (2025-02)
MolSpectra: Pre-training 3D Molecular Representation with Multi-modal Energy Spectra
2602.01585 (2026-02)
A Lightweight Sparse Interaction Network for Time Series Forecasting
and 1 more not named here
MultiHeadAttention
brick of this paper's assembly: entry
1 other paper
1 dated earlier
2510.23577 (2025-10)
TAMI: Taming Heterogeneity in Temporal Interactions for Temporal Graph Link Prediction

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Temporal Graph Neural Networks (TGNNs) aim to capture the evolving structure and timing of interactions in dynamic graphs. Although many models incorporate time through encodings or architectural design, they often compute attention over entangled node and edge representations, failing to reflect their distinct temporal behaviors. Node embeddings evolve slowly as they aggregate long-term structural context, while edge features reflect transient, timestamped interactions (e.g. messages, trades, or transactions). This mismatch results in semantic attention blurring, where attention weights cannot distinguish between slowly drifting node states and rapidly changing, information-rich edge interactions. As a result, models struggle to capture fine-grained temporal dependencies and provide limited transparency into how temporal relevance is computed. This paper introduces KEAT (Kernelized Edge Attention for Temporal Graphs), a novel attention formulation that modulates edge features using a family of continuous-time kernels, including Laplacian, RBF, and learnable MLP variant. KEAT preserves the distinct roles of nodes and edges, and integrates seamlessly with both Transformer-style (e.g., DyGFormer) and message-passing (e.g., TGN) architectures. It achieves up to 18% MRR improvement over the recent DyGFormer and 7% over TGN on link prediction tasks, enabling more accurate, interpretable and temporally aware message passing in TGNNs.

For agents

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

get_harvested_code_for_paper("2602.00596")
get_code_for_paper("2602.00596")
have("2602.00596")

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

curl https://syntology.ai/api/ran/2602.00596.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.00596.svg)](https://syntology.ai/paper/2602.00596)

Connect an agent — have() is free.