SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2111.04840 · ICLR · 2022

Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods

Zhangyang Wang, Sumeet Katariya, Nikhil Rao, Karthik Subbian, Wenqing Zheng, Edward Huang

arXiv · PDF · Open in the Atlas

Code that ran

14 of 16 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, 1 violated it and 10 have no contract verdict; 3 failures of our instrument; 2 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/2111.04840.json · badge: /api/ran/2111.04840.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
amazon-research/gnn-tail-generalization canonical (classifier label)
classifier basis not recorded on the link
14 ran · 2 no recorded run
FunctionStatusWhere it lives
AcontainsB Ran
ran, violated the contract we drafted
amazon-research/gnn-tail-generalization/GNN_model/drop_tricks.py
code served (permissive licence) · get_code("19511295442b8b8a")
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 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 torch

torch.manual_seed(0)

# A can be a string, list, set, or other container that supports 'in' operator
A = "hello world"

# listB should be a list of items that can be checked with 'in' operator against A
listB = ["o", "x", "z"]

ARGS = {
    "A": A,
    "listB": listB
}
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 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
adaptive_auc_loss Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Link_prediction_model/loss.py
code served (permissive licence) · get_code("3cbbe6f92c0de08c")
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 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 torch

torch.manual_seed(0)

# Construct arguments for adaptive_auc_loss
# The function expects:
# - pos_out: tensor of positive scores
# - neg_out: tensor of negative scores
# - num_neg: number of negative samples per positive
# - weight: tensor of weights for each sample

batch_size = 2
num_neg = 3

# pos_out will be reshaped to (-1, 1), so start with shape (2,)
pos_out = torch.randn(batch_size)

# neg_out will be reshaped to (-1, num_neg), so start with shape (batch_size * num_neg,)
neg_out = torch.randn(batch_size * num_neg)

# weight will be reshaped to (-1, 1), so start with shape (batch_size,)
weight = torch.randn(batch_size)

ARGS = {
    "pos_out": pos_out,
    "neg_out": neg_out,
    "num_neg": num_neg,
    "weight": weight,
}
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 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
auc_loss Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Link_prediction_model/loss.py
code served (permissive licence) · get_code("4b9ae549a5732877")
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 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 torch

torch.manual_seed(0)

# The function reshapes pos_out to (-1, 1) and neg_out to (-1, num_neg)
# Then computes element-wise operations and sum
# pos_out should be 1D or flattnable to (batch_size,)
# neg_out should be 1D or flattnable to (batch_size * num_neg,)

batch_size = 2
num_neg = 3

# Create pos_out: should flatten to (batch_size,) -> reshape to (batch_size, 1)
pos_out = torch.randn(batch_size)

# Create neg_out: should flatten to (batch_size * num_neg,) -> reshape to (batch_size, num_neg)
neg_out = torch.randn(batch_size * num_neg)

ARGS = {
    "pos_out": pos_out,
    "neg_out": neg_out,
    "num_neg": num_neg,
}
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 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
build_edge_adj Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Link_prediction_model/edge_LP.py
code served (permissive licence) · get_code("2859a58982511839")
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 collections import defaultdict

torch.manual_seed(0)

# Create a small edge_index tensor
# edge_index should be shape [2, num_edges]
# representing edges between nodes
# Let's create 5 edges between nodes 0-3
edge_index = torch.tensor([
    [0, 1, 2, 1, 0],  # source nodes
    [1, 2, 3, 0, 2]   # target nodes
], dtype=torch.long)

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 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
cal_acc_indices Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/LP_Adj.py
code served (permissive licence) · get_code("2c1fff89f9280558")
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 float, 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 torch

torch.manual_seed(0)

# The function expects:
# - output: a tensor that can be reshaped to 1D, with shape (n,) or (n, 1)
# - labels: a tensor that can be reshaped to 1D

# Create small tensors for testing
output = torch.tensor([[0.8], [0.2], [0.9], [0.1]])  # shape (4, 1)
labels = torch.tensor([1, 0, 1, 0])  # shape (4,)

ARGS = {
    "output": output,
    "labels": labels,
}
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 float, 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
cal_acc_logits Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/LP_Adj.py
code served (permissive licence) · get_code("4995964ae5fddaf1")
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 float, 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 torch

torch.manual_seed(0)

# Create sample logits: batch_size=2, num_classes=4
output = torch.randn(2, 4)

# Create sample labels: batch_size=2
labels = torch.tensor([1, 3])

ARGS = {
    "output": output,
    "labels": labels
}
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 float, 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
diffusion Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/diffusion_feature.py
code served (permissive licence) · get_code("afbdc85089318f81")
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, 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
import numpy as np
from scipy import sparse

torch.manual_seed(0)

# Create small feature matrix (batch_size=2, features=4)
x = np.random.randn(2, 4).astype(np.float64)

# Create small sparse adjacency matrix (2x2)
adj_dense = np.array([[0.9, 0.1], [0.1, 0.9]], dtype=np.float64)
adj = sparse.csr_matrix(adj_dense)

# Number of propagation steps
num_propagations = 3

# Exponent parameter
p = 2.0

# Regularization parameter
alpha = 0.5

ARGS = {
    'x': x,
    'adj': adj,
    'num_propagations': num_propagations,
    'p': p,
    'alpha': alpha
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch, tqdm — 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, 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
gen_normalized_adjs Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/outcome_correlation.py
code served (permissive licence) · get_code("8e4eee31155a52fb")
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 3, 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 numpy as np

torch.manual_seed(0)

# Create a small adjacency matrix (4x4)
# adj should be a 2D tensor that supports element-wise multiplication
adj = torch.randn(4, 4)

# Create D_isqrt vector (4,)
# This represents the inverse square root of the degree matrix diagonal
# It's used in view(-1, 1) and view(1, -1) operations
D_isqrt = torch.randn(4)

ARGS = {
    "adj": adj,
    "D_isqrt": D_isqrt,
}
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 tuple, ndim 3, 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
log_rank_loss Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Link_prediction_model/loss.py
code served (permissive licence) · get_code("03d2b1e423c6445b")
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 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 torch

torch.manual_seed(0)

# The function expects:
# - pos_out: tensor of positive outputs, will be reshaped to (-1, 1)
# - neg_out: tensor of negative outputs, will be reshaped to (-1, num_neg)
# - num_neg: integer specifying number of negative samples

# For pos_out: if we reshape to (-1, 1), we need a flat tensor or any shape that has total elements = batch_size
# For neg_out: if we reshape to (-1, num_neg), we need total elements = batch_size * num_neg
# Let's use batch_size = 2, num_neg = 3

num_neg = 3
batch_size = 2

# pos_out should have batch_size elements (will be reshaped to (batch_size, 1))
pos_out = torch.randn(batch_size)

# neg_out should have batch_size * num_neg elements (will be reshaped to (batch_size, num_neg))
neg_out = torch.randn(batch_size * num_neg)

ARGS = {
    "pos_out": pos_out,
    "neg_out": neg_out,
    "num_neg": num_neg,
}
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 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
reset_weight_GroupNorm Ran
ran; our contract draft was wrong, not the code
amazon-research/gnn-tail-generalization/GNN_model/norm_tricks.py
code served (permissive licence) · get_code("a5483c4f44a5f81a")
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 SimpleArgs. 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

class SimpleArgs:
    def __init__(self):
        self.num_groups = None
        self.miss_rate = 0.5
        self.dataset = 'Cora'
        self.type_model = 'GCN'
        self.num_layers = 4
        self.skip_weight = None
        self.epochs = 100

ARGS = {
    'args': SimpleArgs()
}
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 SimpleArgs — a SimpleArgs output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleArgs 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
run_norm_if_any Ran
ran; our contract draft was wrong, not the code
amazon-research/gnn-tail-generalization/GNN_model/norm_tricks.py
code served (permissive licence) · get_code("429df79768882322")
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, 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
import torch.nn as nn

torch.manual_seed(0)

class MockArgs:
    def __init__(self, type_trick):
        self.type_trick = type_trick

class MockNormLayer(nn.Module):
    def __init__(self):
        super().__init__()
    
    def forward(self, x):
        return x

class MockNet(nn.Module):
    def __init__(self, type_trick='BatchNorm'):
        super().__init__()
        self.args = MockArgs(type_trick=type_trick)
        self.layers_norm = nn.ModuleList([
            MockNormLayer(),
            MockNormLayer(),
            MockNormLayer()
        ])
    
    def forward(self, x):
        return x

net = MockNet(type_trick='BatchNorm')
x = torch.randn(2, 4)
ilayer = 0

ARGS = {
    'net': net,
    'x': x,
    'ilayer': ilayer
}
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, 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
sgc Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/diffusion_feature.py
code served (permissive licence) · get_code("a33fe9e4734902f9")
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 (4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# x is used with @ operator (matrix multiplication with adj)
# It needs to be a numpy array (since adj @ x is performed and result is converted from numpy)
x = np.random.randn(4, 3).astype(np.float32)

# adj is used with @ operator (matrix multiplication with x)
# It needs to be a numpy array that can multiply with x
# Shape should be (4, 4) to multiply with (4, 3)
adj = np.random.randn(4, 4).astype(np.float32)

# num_propagations is used in range(), so it's an integer
num_propagations = 2

ARGS = {
    "x": x,
    "adj": adj,
    "num_propagations": num_propagations,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch, tqdm — 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 (4, 3) — shape (4, 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
to_device Ran
ran on a synthesized input
amazon-research/gnn-tail-generalization/Label_propagation_model/LP_Adj.py
code served (permissive licence) · get_code("bfdd8eb94daad37e")
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
import torch.nn as nn

torch.manual_seed(0)

# Create a list with mixed content: dicts containing tensors and tensors
list1d = [
    {
        'weight': torch.randn(2, 4),
        'bias': torch.randn(3)
    },
    torch.randn(2, 5),
    {
        'data': torch.randn(4, 3)
    }
]

device = 'cpu'

ARGS = {
    'list1d': list1d,
    'device': device
}
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
tonp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
amazon-research/gnn-tail-generalization/GNN_model/GCN.py
code served (permissive licence) · get_code("18d569c66cbe1eda")
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 ndarray, ndim 2, dtype family float, 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 as th
import numpy as np

th.manual_seed(0)

# The function tonp takes one parameter 'arr' which can be either a torch.Tensor or something convertible to numpy array
# It checks if arr is a th.Tensor, and if so calls detach().cpu().data.numpy() on it
# Otherwise it calls np.asarray on it

# Test case 1: a torch.Tensor
tensor_arg = th.randn(2, 4)

ARGS = {
    "arr": tensor_arg
}
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 ndarray, ndim 2, dtype family float, 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
general_outcome_correlation_YAG Not yet run
unverified (harvested, no recorded run)
amazon-research/gnn-tail-generalization/Label_propagation_model/outcome_correlation.py
code served (permissive licence) · get_code("1c9f7749bb1f3502")
How it ran

no invocation recorded.

lp Not yet run
unverified (harvested, no recorded run)
amazon-research/gnn-tail-generalization/Label_propagation_model/diffusion_feature.py
code served (permissive licence) · get_code("869350b6e16a5a33")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 12 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
AcontainsB 2 other papers
2 later
2308.06801 (2023-08)
SAILOR: Structural Augmentation Based Tail Node Representation Learning
2406.09870 (2024-06)
IGL-Bench: Establishing the Comprehensive Benchmark for Imbalanced Graph Learning
reset_weight_GroupNorm 2 other papers
2 later
2308.06801 (2023-08)
SAILOR: Structural Augmentation Based Tail Node Representation Learning
2406.09870 (2024-06)
IGL-Bench: Establishing the Comprehensive Benchmark for Imbalanced Graph Learning
run_norm_if_any 2 other papers
2 later
2308.06801 (2023-08)
SAILOR: Structural Augmentation Based Tail Node Representation Learning
2406.09870 (2024-06)
IGL-Bench: Establishing the Comprehensive Benchmark for Imbalanced Graph Learning
tonp 1 other paper
1 later
2406.09870 (2024-06)
IGL-Bench: Establishing the Comprehensive Benchmark for Imbalanced Graph Learning

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

Graph Neural Networks (GNNs) have achieved state-of-the-art performance in node classification, regression, and recommendation tasks. GNNs work well when rich and high-quality connections are available. However, their effectiveness is often jeopardized in many real-world graphs in which node degrees have power-law distributions. The extreme case of this situation, where a node may have no neighbors, is called Strict Cold Start (SCS). SCS forces the prediction to rely completely on the node's own features. We propose Cold Brew, a teacher-student distillation approach to address the SCS and noisy-neighbor challenges for GNNs. We also introduce feature contribution ratio (FCR), a metric to quantify the behavior of inductive GNNs to solve SCS. We experimentally show that FCR disentangles the contributions of different graph data components and helps select the best architecture for SCS generalization. We further demonstrate the superior performance of Cold Brew on several public benchmark and proprietary e-commerce datasets, where many nodes have either very few or noisy connections. Our source code is available at https: //github.com/amazon-research/gnn-tail-generalization.

For agents

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

get_harvested_code_for_paper("2111.04840")
get_code_for_paper("2111.04840")
have("2111.04840")

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

curl https://syntology.ai/api/ran/2111.04840.json

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

[![Syntology run record](https://syntology.ai/api/ran/2111.04840.svg)](https://syntology.ai/paper/2111.04840)

Connect an agent — have() is free.