SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.08683 · IJCAI · 2021

Learning Embeddings from Knowledge Graphs With Numeric Edge Attributes

Sumit Pai, Luca Costabello

arXiv · PDF · Open in the Atlas

Code that ran

11 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 failures of our instrument; 5 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/2105.08683.json · badge: /api/ran/2105.08683.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
Accenture/AmpliGraph canonical (classifier label)
classifier basis: census
11 ran · 5 no recorded run
FunctionStatusWhere it lives
clip_before_exp Ran
ran, honoured the contract we drafted
Accenture/AmpliGraph/ampligraph/latent_features/loss_functions.py
code served (permissive licence) · get_code("73cdfc67eab82a51")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

DEFAULT_CLIP_EXP_LOWER = -75.0
DEFAULT_CLIP_EXP_UPPER = 75.0

def clip_before_exp(value):
    """Clip the value for stability of exponential."""
    return tf.clip_by_value(value, clip_value_min=DEFAULT_CLIP_EXP_LOWER, clip_value_max=DEFAULT_CLIP_EXP_UPPER)

# Create a small TensorFlow tensor with values that need clipping
# Mix of values below, within, and above the clipping range
value_data = np.array([[-100.0, -50.0, 0.0, 50.0],
                       [100.0, -75.5, 75.5, 25.0]], dtype=np.float32)
value_tensor = tf.constant(value_data)

ARGS = {
    "value": value_tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
dataframe_to_triples Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/utils/model_utils.py
code served (permissive licence) · get_code("1e631150b0c80c72")
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 (9, 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 pandas as pd
import numpy as np

# Create a small DataFrame with some columns
X = pd.DataFrame({
    'species': ['setosa', 'versicolor', 'virginica'],
    'sepal_length': [5.1, 7.0, 6.3],
    'petal_width': [0.2, 1.4, 1.9],
    'color': ['red', 'blue', 'green']
})

# Define schema as a list of tuples (subject, relation_name, object)
# where subject and object are column names in X
schema = [
    ['species', 'has_sepal_length', 'sepal_length'],
    ['species', 'has_petal_width', 'petal_width'],
    ['color', 'belongs_to_species', 'species']
]

ARGS = {
    'X': X,
    'schema': schema
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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 (9, 3) — shape (9, 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_number_of_partitions Ran
ran, honoured the contract we drafted
Accenture/AmpliGraph/ampligraph/datasets/graph_partitioner.py
code served (permissive licence) · get_code("eeee235b35bf4027")
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 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 numpy as np

# The function get_number_of_partitions takes a single parameter n (int)
# and computes the number of partitions using the formula: n * (n + 1) / 2
# This is a simple mathematical function with no complex dependencies.

# Define arguments for the function
ARGS = {
    "n": 5
}
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 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
load_from_csv Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/datasets/datasets.py
code served (permissive licence) · get_code("e71579ce86730827")
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 (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 os
import tempfile
import pandas as pd

# Create a temporary directory and a test CSV file
temp_dir = tempfile.mkdtemp()
csv_file = os.path.join(temp_dir, 'test_triples.csv')

# Create a simple CSV file with triples (subject, relation, object)
triples_data = """entity1	relation1	entity2
entity2	relation2	entity3
entity1	relation3	entity3
entity2	relation1	entity1"""

with open(csv_file, 'w') as f:
    f.write(triples_data)

ARGS = {
    'directory_path': temp_dir,
    'file_name': 'test_triples.csv',
    'sep': '\t',
    'header': None,
    'add_reciprocal_rels': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — 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 (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
load_mapper_from_json Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/datasets/datasets.py
code served (permissive licence) · get_code("b7720ad19371898e")
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 dict. 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 json
import os
import tempfile

# Create a temporary directory and a valid mapper JSON file
temp_dir = tempfile.mkdtemp()
mapper_data = {
    "/m/234fsd/": "Dog",
    "/m/567ghi/": "Cat",
    "relation_has_color": "has_color",
    "color_black": "Black"
}

mapper_file = "mapper.json"
mapper_path = os.path.join(temp_dir, mapper_file)
with open(mapper_path, 'w') as f:
    json.dump(mapper_data, f)

ARGS = {
    "directory_path": temp_dir,
    "file_name": mapper_file
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
preprocess_focusE_weights Ran
ran, honoured the contract we drafted
Accenture/AmpliGraph/ampligraph/utils/model_utils.py
code served (permissive licence) · get_code("4d2cf14a78b063c4")
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 (6, 1) 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 numpy as np
import pandas as pd

# Create sample data and weights
# data should have shape (n, 3) or (n, 4) with triples and optionally weights
# The function uses data[:, 1] to extract relations
np.random.seed(0)

# Create sample data with 6 triples and 3 columns (entity, relation, entity)
n_samples = 6
data = np.array([
    [0, 0, 1],
    [1, 0, 2],
    [2, 1, 3],
    [3, 1, 4],
    [4, 2, 5],
    [5, 2, 0]
], dtype=np.float32)

# Create sample weights with shape (n, 1) or (n, m) where m > 1
# Mix in some NaN values to test the NaN handling
weights = np.array([
    [0.1],
    [0.5],
    [np.nan],
    [0.8],
    [0.3],
    [0.7]
], dtype=np.float32)

# Alternative: test with multiple weight columns
# weights = np.array([
#     [0.1, 0.2],
#     [0.5, 0.6],
#     [np.nan, 0.3],
#     [0.8, 0.9],
#     [0.3, 0.4],
#     [0.7, 0.8]
# ], dtype=np.float32)

ARGS = {
    "data": data,
    "weights": weights,
    "normalize": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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 (6, 1) — shape (6, 1) 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
register_compatibility Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/compat/models.py
code served (permissive licence) · get_code("6f8298f9f8bd8ad0")
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 function. 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)

# The function register_compatibility takes one parameter: 'name'
# It returns a decorator that takes a class_handle and registers it
# The decorator sets a 'name' attribute on the class and returns it

# To test this function, we need to provide a valid name string
# and optionally use the decorator on a test class

# Create a simple test class that could be decorated
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

# Define the arguments for register_compatibility
ARGS = {
    "name": "test_model_v1"
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
register_indexer_backend Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/datasets/data_indexer.py
code served (permissive licence) · get_code("79be7853f34faf64")
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 function. 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 logging

logger = logging.getLogger(__name__)

# Clear the registry before testing
INDEXER_BACKEND_REGISTRY = {}

def register_indexer_backend(name):
    """Decorator responsible for registering partition in the partition registry.

    Parameters
    ----------
    name: str
         Name of the new backend.

    Example
    -------
    >>>@register_indexer_backend("NewBackendName")
    >>>class NewBackend():
    >>>... pass
    """

    def insert_in_registry(class_handle):
        """Checks if backend already exists and if not registers it."""
        if name in INDEXER_BACKEND_REGISTRY.keys():
            msg = 'Indexer backend with name {} already exists!'.format(name)
            logger.error(msg)
            raise Exception(msg)
        INDEXER_BACKEND_REGISTRY[name] = class_handle
        class_handle.name = name
        return class_handle
    return insert_in_registry


# Create a simple test backend class
class TestBackend:
    """A simple test backend class."""
    pass


# Define the arguments for the function
ARGS = {
    "name": "TestBackendName"
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
register_loss Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/latent_features/loss_functions.py
code served (permissive licence) · get_code("74d4214ef4c6a401")
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 function. 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 simple test class to use as a loss function
class SimpleLoss(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 2)
    
    def forward(self, x, y):
        return self.linear(x).mean()

# Define arguments for register_loss function
ARGS = {
    "name": "mse_loss",
    "external_params": ["learning_rate", "weight_decay"]
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
register_partitioning_manager Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/datasets/partitioned_data_manager.py
code served (permissive licence) · get_code("c655ffc4d09dc380")
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 function. 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 logging

logger = logging.getLogger(__name__)

PARTITION_MANAGER_REGISTRY = {}

def register_partitioning_manager(name):
    """Decorator responsible for registering partition manager in the partition manager registry.

    Parameters
    ----------
    name: str
         Name of the new partition manager.

    Example
    -------
    >>>@register_partitioning_manager("NewManagerName")
    >>>class NewManagerName(PartitionDataManager):
    >>>... pass
    """

    def insert_in_registry(class_handle):
        """Checks if partition manager already exists and if not registers it."""
        if name in PARTITION_MANAGER_REGISTRY.keys():
            msg = 'Partitioning Manager with name {} already exists!'.format(name)
            logger.error(msg)
            raise Exception(msg)
        PARTITION_MANAGER_REGISTRY[name] = class_handle
        class_handle.name = name
        return class_handle
    return insert_in_registry


class PartitionDataManager:
    """Base class for partition data managers."""
    pass


class TestManager(PartitionDataManager):
    """Test partition manager class."""
    pass


ARGS = {
    "name": "TestPartitionManager"
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
register_partitioning_strategy Ran
ran; our contract draft was wrong, not the code
Accenture/AmpliGraph/ampligraph/datasets/graph_partitioner.py
code served (permissive licence) · get_code("f03691b5daa8b1aa")
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 function. 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 logging

logger = logging.getLogger(__name__)

PARTITION_ALGO_REGISTRY = {}

def register_partitioning_strategy(name, manager):
    """Decorator responsible for registering partition in the partition registry.

    Parameters
    ----------
    name: str
         Name of the new partition strategy.
    manager: str
         Name of the partitioning manager that will handle this partitioning strategy during training.

    Example
    -------
    >>>@register_partitioning_strategy("NewStrategyName")
    >>>class NewPartitionStrategy(AbstractGraphPartitioner):
    >>>... pass
    """

    def insert_in_registry(class_handle):
        """Checks if partition already exists and if not registers it."""
        if name in PARTITION_ALGO_REGISTRY.keys():
            msg = 'Partitioning Strategy with name {} already exists!'.format(name)
            logger.error(msg)
            raise Exception(msg)
        PARTITION_ALGO_REGISTRY[name] = class_handle
        class_handle.name = name
        class_handle.manager = manager
        return class_handle
    return insert_in_registry


class MockPartitionStrategy:
    """Mock partitioning strategy class for testing."""
    pass


ARGS = {
    "name": "TestPartitionStrategy",
    "manager": "DefaultPartitionManager"
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
compare_partitionings Not yet run
unverified (harvested, no recorded run)
Accenture/AmpliGraph/ampligraph/datasets/partitioning_reporter.py
code served (permissive licence) · get_code("e6b2d5013e134fbd")
How it ran

no invocation recorded.

evaluate_performance Not yet run
unverified (harvested, no recorded run)
Accenture/AmpliGraph/ampligraph/compat/evaluate.py
code served (permissive licence) · get_code("b2164845d6c66f4a")
How it ran

no invocation recorded.

get Not yet run
unverified (harvested, no recorded run)
Accenture/AmpliGraph/ampligraph/latent_features/loss_functions.py
code served (permissive licence) · get_code("e5cdd5cb67b2cb72")
How it ran

no invocation recorded.

get_partition_adapter Not yet run
unverified (harvested, no recorded run)
Accenture/AmpliGraph/ampligraph/datasets/partitioned_data_manager.py
code served (permissive licence) · get_code("44fdbf9bdcea6697")
How it ran

no invocation recorded.

restore_model Not yet run
unverified (harvested, no recorded run)
Accenture/AmpliGraph/ampligraph/utils/model_utils.py
code served (permissive licence) · get_code("eed65ab60f7570f8")
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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 16 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.

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

Numeric values associated to edges of a knowledge graph have been used to represent uncertainty, edge importance, and even out-of-band knowledge in a growing number of scenarios, ranging from genetic data to social networks. Nevertheless, traditional knowledge graph embedding models are not designed to capture such information, to the detriment of predictive power. We propose a novel method that injects numeric edge attributes into the scoring layer of a traditional knowledge graph embedding architecture. Experiments with publicly available numeric-enriched knowledge graphs show that our method outperforms traditional numeric-unaware baselines as well as the recent UKGE model.

For agents

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

get_harvested_code_for_paper("2105.08683")
get_code_for_paper("2105.08683")
have("2105.08683")

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

curl https://syntology.ai/api/ran/2105.08683.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.08683.svg)](https://syntology.ai/paper/2105.08683)

Connect an agent — have() is free.