SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2010.02803 · 2020

A Transformer-based Framework for Multivariate Time Series Representation Learning

arXiv · PDF · Open in the Atlas

Code that ran

23 of 32 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; 1 honoured the contract we drafted, 0 violated it and 15 have no contract verdict; 7 failures of our instrument; 9 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2010.02803.json · badge: /api/ran/2010.02803.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
gzerveas/mvts_transformer canonical (classifier label)
classifier basis: census
16 ran · 4 no recorded run
appleparan/mise.py community (classifier label)
classifier basis not recorded on the link
6 ran · 4 no recorded run
pkuyym/evolvingattention extension (classifier label)
classifier basis: classifier_2025
1 ran · 1 no recorded run
FunctionStatusWhere it lives
acc_top_k Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/utils/analysis.py
code served (permissive licence) · get_code("4f27929df483e1c4")
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 1, dtype family float, shape (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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create sample data
# predictions: (N_samples, k) array of top class indices
# Let's use N_samples=2, k=3 (top 3 predictions per sample)
N_samples = 2
k = 3
num_classes = 5

# Generate predictions: top k class indices for each sample (pre-sorted by score)
# Shape: (2, 3)
predictions = np.array([
    [0, 1, 2],  # sample 0: top 3 predictions are classes 0, 1, 2
    [2, 3, 4]   # sample 1: top 3 predictions are classes 2, 3, 4
])

# Generate ground truth labels
# Shape: (2,)
y_true = np.array([1, 2])

ARGS = {
    'predictions': predictions,
    'y_true': y_true
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (3,) — shape (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
accuracy Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/utils/analysis.py
code served (permissive licence) · get_code("431078d6fcc08958")
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 float64, ndim 0, 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 numpy as np

# Create simple numpy arrays for predictions and ground truth
y_pred = np.array([0, 1, 2, 1, 0, 2, 1, 0])
y_true = np.array([0, 1, 2, 1, 0, 1, 1, 0])

# Define excluded labels as a list
excluded_labels = [2]

ARGS = {
    "y_pred": y_pred,
    "y_true": y_true,
    "excluded_labels": excluded_labels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
collate_superv Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/datasets/dataset.py
code served (permissive licence) · get_code("94120ad91132bf72")
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

torch.manual_seed(0)

# Create sample data: list of (features, labels, IDs) tuples
# features: (seq_length, feat_dim) tensor
# labels: (num_labels,) tensor
# IDs: some identifier

batch_size = 2
feat_dim = 4
num_labels = 3

data = []
for i in range(batch_size):
    seq_length = 5 + i  # variable sequence lengths: 5, 6
    features = torch.randn(seq_length, feat_dim)
    labels = torch.randint(0, 2, (num_labels,)).float()
    ID = f"sample_{i}"
    data.append((features, labels, ID))

ARGS = {
    "data": data,
    "max_len": None
}
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
compensate_masking Ran
ran; our contract draft was wrong, not the code
gzerveas/mvts_transformer/src/datasets/dataset.py
code served (permissive licence) · get_code("0c1e47b7a9f0bc5c")
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 3, dtype family float, shape (2, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for testing
batch_size = 2
seq_length = 3
feat_dim = 4

# X: (batch_size, seq_length, feat_dim) - the feature vectors
X = torch.randn(batch_size, seq_length, feat_dim)

# mask: (batch_size, seq_length, feat_dim) - binary mask where 1 = active, 0 = masked
# Create a mask with some elements masked (0) and some active (1)
mask = torch.tensor([
    [[1, 1, 0, 1],
     [1, 0, 1, 1],
     [0, 1, 1, 1]],
    [[1, 1, 1, 0],
     [1, 1, 0, 1],
     [1, 0, 1, 1]]
], dtype=torch.float32)

ARGS = {
    "X": X,
    "mask": mask
}
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
convert_metrics_per_batch_to_per_sample Ran
ran; our contract draft was wrong, not the code
gzerveas/mvts_transformer/src/running.py
pointer only (licence: MIT); the stored copy get_code serves, from DAMO-DI-ML/One_Fits_All, is pointer only (licence: NONE) · get_code("f79cf1a40c083fee")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from DAMO-DI-ML/One_Fits_All, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_loss_module Ran
ran; our contract draft was wrong, not the code
gzerveas/mvts_transformer/src/models/loss.py
pointer only (licence: MIT); the stored copy get_code serves, from scxsunchenxi/test, is pointer only (licence: NONE) · get_code("5346622dac785f32")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from scxsunchenxi/test, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_optimizer Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/optimizers.py
code served (permissive licence) · get_code("98a9799efb4a305c")
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 type. 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
from torch.optim.optimizer import Optimizer
import math

torch.manual_seed(0)

# Create a simple model with parameters
model = nn.Linear(4, 3)

# Get the parameters from the model
params = model.parameters()

# Define ARGS dictionary with all parameters for get_optimizer
ARGS = {
    'name': 'RAdam'
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
get_pos_encoder Ran
ran; our contract draft was wrong, not the code
pkuyym/evolvingattention/EA_DC_Transformer/src/models/ea_dc_transformer.py
pointer only (licence: NONE) · get_code("578b29c746156148")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, pkuyym/evolvingattention 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.

interpolate_missing Ran
ran; our contract draft was wrong, not the code
gzerveas/mvts_transformer/src/datasets/data.py
code served (permissive licence) · get_code("c43495731acfe20c")
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 Series. 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 pandas as pd
import numpy as np

# Create a pandas Series with some NaN values for testing
y = pd.Series([1.0, np.nan, 3.0, np.nan, 5.0, 6.0])

ARGS = {
    'y': y
}
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 Series — a Series output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Series output the recorded ndim 1 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
l2_reg_loss Ran
ran, honoured the contract we drafted
gzerveas/mvts_transformer/src/models/loss.py
pointer only (licence: MIT); the stored copy get_code serves, from scxsunchenxi/test, is pointer only (licence: NONE) · get_code("96c6fa7dd49845b2")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from scxsunchenxi/test, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

load Ran
ran on a synthesized input
appleparan/mise.py/mise/data.py
code served (permissive licence) · get_code("f766b8a4595da2de")
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 DataFrame. 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 pandas as pd
import io
import tempfile
import os

# Create a temporary CSV file with sample data
csv_content = """idx1,idx2,date_col,value
A,X,2023-01-01,100
A,Y,2023-01-02,200
B,X,2023-01-03,300
B,Y,2023-01-04,400
C,X,2023-01-05,500
C,Y,2023-01-06,600
D,X,2023-01-07,700
D,Y,2023-01-08,800
E,X,2023-01-09,900
E,Y,2023-01-10,1000
F,X,2023-01-11,1100
F,Y,2023-01-12,1200
"""

# Write to a temporary file
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.csv', delete=False)
temp_file.write(csv_content)
temp_file.close()

ARGS = {
    'datecol': [2],
    'filepath': temp_file.name
}
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 DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
load_config Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/utils/utils.py
code served (permissive licence) · get_code("db289815c50136e9")
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 tempfile
import os

# Create a temporary config file with valid JSON content
config_data = {
    "model": {
        "name": "test_model",
        "layers": 3,
        "hidden_dim": 128
    },
    "training": {
        "batch_size": 32,
        "learning_rate": 0.001,
        "epochs": 10
    },
    "data": {
        "train_path": "/path/to/train",
        "val_path": "/path/to/val",
        "num_classes": 10
    }
}

# Write config to a temporary file
temp_dir = tempfile.gettempdir()
config_filepath = os.path.join(temp_dir, "test_config.json")
with open(config_filepath, 'w') as f:
    json.dump(config_data, f)

ARGS = {
    "config_filepath": config_filepath
}
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
load_imputed Ran
ran on a synthesized input
appleparan/mise.py/mise/data.py
code served (permissive licence) · get_code("e3c61c91b91a7e4e")
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 DataFrame. 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 pandas as pd
import tempfile
import os

# Create a temporary CSV file with sample data
temp_dir = tempfile.mkdtemp()
filepath = os.path.join(temp_dir, 'input.csv')

# Create sample CSV data with multi-index (first two columns) and a date column
csv_data = """idx1,idx2,date,value
a,1,2023-01-01,100
a,2,2023-01-02,200
b,1,2023-01-03,300
b,2,2023-01-04,400"""

with open(filepath, 'w') as f:
    f.write(csv_data)

# Define arguments
ARGS = {
    'datecol': [2],  # Column index 2 contains dates
    'filepath': filepath
}
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 DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
load_station Ran
ran on a synthesized input
appleparan/mise.py/mise/data.py
code served (permissive licence) · get_code("622e8059e9d2a770")
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 DataFrame. 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 pandas as pd
import numpy as np

# Create a sample DataFrame with station codes
data = {
    'stationCode': [111123, 111124, 111125, 111123, 111126],
    'temperature': [20.5, 21.3, 19.8, 22.1, 20.0],
    'humidity': [65, 72, 58, 70, 68],
    'pressure': [1013.25, 1012.80, 1014.10, 1013.50, 1012.90]
}
df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'code': 111123
}
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 DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
parse_key Ran
ran on a synthesized input
appleparan/mise.py/mise/utils.py
code served (permissive licence) · get_code("bf6fb185b4d4efb5")
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, ndim 1, 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 pandas as pd
import numpy as np
from datetime import datetime, timedelta

# Create a small DataFrame with DateTimeIndex
dates = pd.date_range(start='2023-01-15', periods=2, freq='D')
df = pd.DataFrame(
    {'value': [1.0, 2.0]},
    index=dates
)

ARGS = {
    'df': df,
    'period': 'y'
}
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 list, ndim 1, dtype family float — a list 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
parse_ykey Ran
ran on a synthesized input
appleparan/mise.py/mise/utils.py
code served (permissive licence) · get_code("068024601e1b87e3")
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 str. 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 pandas as pd

# Create a DatetimeIndex with a specific date
idx = pd.DatetimeIndex(['2023-03-15'])

ARGS = {
    'idx': idx
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
precision Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/utils/analysis.py
code served (permissive licence) · get_code("839af27815f6fa38")
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 float64, ndim 0, 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 numpy as np

# The function takes:
# - y_true: array of true labels
# - y_pred: array of predicted labels
# - label: the class index to compute precision for

# Create small numpy arrays for true and predicted labels
y_true = np.array([0, 1, 1, 0, 1, 2, 2, 0])
y_pred = np.array([0, 1, 0, 0, 1, 2, 1, 0])
label = 1

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
    "label": label
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
process_data Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/datasets/utils.py
code served (permissive licence) · get_code("51e3a59616ee190e")
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 3, dtype family float, shape (2, 5, 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 numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from tqdm import tqdm

# Create mock data structure similar to a pandas DataFrame
# The function expects X.iloc[i, :] to return a Series-like object
# where each element is a Series with .isnull(), .interpolate(), and len()

class MockSeries:
    def __init__(self, data):
        self.data = np.array(data, dtype=float)
    
    def isnull(self):
        class NullChecker:
            def any(self):
                return False
        return NullChecker()
    
    def interpolate(self, method='linear', limit_direction='both'):
        return self
    
    def __len__(self):
        return len(self.data)
    
    def __iter__(self):
        return iter(self.data)

class MockDataFrame:
    def __init__(self, data):
        self.data = data
    
    def __len__(self):
        return len(self.data)
    
    def iloc(self, i, slice_obj):
        # Returns a list of MockSeries objects
        class ILocIndexer:
            def __init__(self, parent):
                self.parent = parent
            
            def __getitem__(self, key):
                i = key[0]
                return self.parent.data[i]
        
        return ILocIndexer(self)

# Create sample data: 2 rows, each row has 3 time series of varying lengths
row1 = [
    MockSeries([1.0, 2.0, 3.0, 4.0, 5.0]),
    MockSeries([2.0, 4.0, 6.0]),
    MockSeries([1.0, 1.5, 2.0, 2.5, 3.0, 3.5])
]

row2 = [
    MockSeries([5.0, 4.0, 3.0, 2.0]),
    MockSeries([1.0, 2.0, 3.0, 4.0, 5.0, 6.0]),
    MockSeries([10.0, 20.0, 30.0])
]

mock_df_data = [row1, row2]

# Create a proper pandas DataFrame-like mock
X = pd.DataFrame({
    'col1': [pd.Series([1.0, 2.0, 3.0, 4.0, 5.0]), pd.Series([5.0, 4.0, 3.0, 2.0])],
    'col2': [pd.Series([2.0, 4.0, 6.0]), pd.Series([1.0, 2.0, 3.0, 4.0, 5.0, 6.0])],
    'col3': [pd.Series([1.0, 1.5, 2.0, 2.5, 3.0, 3.5]), pd.Series([10.0, 20.0, 30.0])]
})

ARGS = {
    'X': X,
    'min_len': 5,
    'normalise': 'standard'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 5, 3) — shape (2, 5, 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
subsample Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gzerveas/mvts_transformer/src/datasets/data.py
pointer only (licence: MIT); the stored copy get_code serves, from scxsunchenxi/test, is pointer only (licence: NONE) · get_code("99cb2d7796fc399b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from scxsunchenxi/test, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

swish Ran
ran on a synthesized input
appleparan/mise.py/mise/dl/rnn_mul_lstnet_skip.py
code served (permissive licence) · get_code("048af2894f1a47f8")
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

torch.manual_seed(0)

_input = torch.randn(2, 4)
beta = 1.0

ARGS = {
    "_input": _input,
    "beta": beta,
}
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
timer Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/utils/utils.py
code served (permissive licence) · get_code("8269a8d00bc80c97")
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 functools
import time

def timer(func):
    """Print the runtime of the decorated function"""

    @functools.wraps(func)
    def wrapper_timer(*args, **kwargs):
        start_time = time.perf_counter()
        value = func(*args, **kwargs)
        end_time = time.perf_counter()
        run_time = end_time - start_time
        print(f'Finished {func.__name__!r} in {run_time} secs')
        return value
    return wrapper_timer


def example_function(x, y):
    """A simple function to be decorated"""
    return x + y


ARGS = {
    'func': example_function
}
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
transduct_mask Ran
ran on a synthesized input
gzerveas/mvts_transformer/src/datasets/dataset.py
code served (permissive licence) · get_code("f77dc925dc416cc7")
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 (5, 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 numpy as np

# Create a small (seq_length=5, feat_dim=3) numpy array
X = np.random.randn(5, 3).astype(np.float32)

# List of feature indices to mask
mask_feats = [0, 2]

# Proportion at the start to skip masking
start_hint = 0.2

# Proportion at the end to skip masking
end_hint = 0.1

ARGS = {
    "X": X,
    "mask_feats": mask_feats,
    "start_hint": start_hint,
    "end_hint": end_hint,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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 (5, 3) — shape (5, 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
uniform_scaling Ran
ran; our contract draft was wrong, not the code
gzerveas/mvts_transformer/src/datasets/utils.py
code served (permissive licence) · get_code("6c254811cc3193b1")
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, ndim 1, 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)

# Analyze the function:
# - data: needs to support len() and indexing with int
# - max_len: an integer specifying the target length
# The function samples from data uniformly across its length

# Create a simple time series as a list
data = [1.0, 2.5, 3.7, 4.2, 5.1, 6.0, 7.3, 8.1]
max_len = 5

ARGS = {
    "data": data,
    "max_len": max_len,
}
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 list, ndim 1, dtype family float — a list 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
dataset2svinput Not yet run
unverified (harvested, no recorded run)
appleparan/mise.py/mise/dl/dt_xgboost.py
code served (permissive licence) · get_code("c600f54d66a899c1")
How it ran

no invocation recorded.

load_from_tsfile_to_dataframe Not yet run
unverified (harvested, no recorded run)
gzerveas/mvts_transformer/src/datasets/utils.py
code served (permissive licence) · get_code("f88f4ea73750eaa3")
How it ran

no invocation recorded.

load_model Not yet run
unverified (harvested, no recorded run)
gzerveas/mvts_transformer/src/utils/utils.py
code served (permissive licence) · get_code("06f548c9975c99a4")
How it ran

no invocation recorded.

model_factory Not yet run
unverified (harvested, no recorded run)
pkuyym/evolvingattention/EA_DC_Transformer/src/models/ea_dc_transformer.py
pointer only (licence: NONE) · get_code("1dac9a2dac1bfabb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, pkuyym/evolvingattention 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.

model_factory Not yet run
unverified (harvested, no recorded run)
gzerveas/mvts_transformer/src/models/ts_transformer.py
code served (permissive licence) · get_code("131b5708bcdc85d1")
How it ran

no invocation recorded.

mw_df Not yet run
unverified (harvested, no recorded run)
appleparan/mise.py/mise/dl/dt_xgboost.py
code served (permissive licence) · get_code("935882a101552616")
How it ran

no invocation recorded.

parse_wkey Not yet run
unverified (harvested, no recorded run)
appleparan/mise.py/mise/utils.py
code served (permissive licence) · get_code("8bbe4bc623baaf81")
How it ran

no invocation recorded.

sim_xgboost Not yet run
unverified (harvested, no recorded run)
appleparan/mise.py/mise/dl/dt_xgboost.py
code served (permissive licence) · get_code("f283f0737d07a747")
How it ran

no invocation recorded.

split_dataset Not yet run
unverified (harvested, no recorded run)
gzerveas/mvts_transformer/src/datasets/datasplit.py
code served (permissive licence) · get_code("761d7abb12b095eb")
How it ran

no invocation recorded.

Also present in other papers' code

Of 32 samples lifted from this paper's linked repositories, 32 carry a function identity and 0 do not. Of the 32 distinct functions with an identity, 8 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 8 later) and 24 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
convert_metrics_per_batch_to_per_sample 3 other papers
3 later
2308.08241 (2023-08)
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
2409.17515 (2024-09)
From News to Forecast: Integrating Event Analysis in LLM-Based Time Series Forecasting with Reflection
2510.19710 (2025-10)
SEMPO: Lightweight Foundation Models for Time Series Forecasting
interpolate_missing 3 other papers
3 later
2308.08241 (2023-08)
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
2403.07300 (2024-03)
CALF: Aligning LLMs for Time Series Forecasting via Cross-modal Fine-Tuning
2502.06910 (2025-02)
TimeKAN: KAN-based Frequency Decomposition Learning Architecture for Long-term Time Series Forecasting
compensate_masking 2 other papers
2 later
2411.07506 (2024-11)
FlowTS: Time Series Generation via Rectified Flow
2412.09328 (2024-12)
Auto-Regressive Moving Diffusion Models for Time Series Forecasting
get_pos_encoder 2 other papers
2 later
2202.06258 (2022-02)
Flowformer: Linearizing Transformers with Conservation Flows
2410.17918 (2024-10)
Addressing Asynchronicity in Clinical Multimodal Fusion via Individualized Chest X-ray Generation
get_loss_module 1 other paper
1 later
2308.08241 (2023-08)
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
l2_reg_loss 1 other paper
1 later
2308.08241 (2023-08)
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
subsample 1 other paper
1 later
2308.08241 (2023-08)
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
uniform_scaling 1 other paper
1 later
2310.05467 (2023-10)
Temporal Convolutional Explorer Helps Understand 1D-CNN's Learning Behavior in Time Series Classification from Frequency Domain

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

In this work we propose for the first time a transformer-based framework for unsupervised representation learning of multivariate time series. Pre-trained models can be potentially used for downstream tasks such as regression and classification, forecasting and missing value imputation. By evaluating our models on several benchmark datasets for multivariate time series regression and classification, we show that not only does our modeling approach represent the most successful method employing unsupervised learning of multivariate time series presented to date, but also that it exceeds the current state-of-the-art performance of supervised methods; it does so even when the number of training samples is very limited, while offering computational efficiency. Finally, we demonstrate that unsupervised pre-training of our transformer models offers a substantial performance benefit over fully supervised learning, even without leveraging additional unlabeled data, i.e., by reusing the same data samples through the unsupervised objective.

For agents

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

get_harvested_code_for_paper("2010.02803")
get_code_for_paper("2010.02803")
have("2010.02803")

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

curl https://syntology.ai/api/ran/2010.02803.json

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

[![Syntology run record](https://syntology.ai/api/ran/2010.02803.svg)](https://syntology.ai/paper/2010.02803)

Connect an agent — have() is free.