SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2007.02871 · 2020

DART: Open-Domain Structured Data Record to Text Generation

arXiv · PDF · Open in the Atlas

Code that ran

10 of 14 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, 0 violated it and 10 have no contract verdict; 0 failures of our instrument; 4 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/2007.02871.json · badge: /api/ran/2007.02871.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
Yale-LILY/dart canonical (classifier label)
classifier basis: census
2 ran · 0 no recorded run
Mandar-Sharma/TCube community (classifier label)
classifier basis not recorded on the link
8 ran · 4 no recorded run
FunctionStatusWhere it lives
best_line Ran
ran on a synthesized input
Mandar-Sharma/TCube/segmentation.py
code served (permissive licence) · get_code("3e9d5f78e87be8cf")
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 (6,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct arguments for best_line function
# The function takes:
# - remaining: array-like of numeric values (used with slicing, np.append, np.polyfit)
# - max_error: numeric threshold for error comparison

remaining = np.array([1.0, 2.1, 3.05, 4.2, 5.1, 6.0])
max_error = 0.5

ARGS = {
    "remaining": remaining,
    "max_error": max_error
}
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 (6,) — shape (6,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_graph_covid Ran
ran on a synthesized input
Mandar-Sharma/TCube/data2graph.py
code served (permissive licence) · get_code("5d64cb47034e16ff")
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 pandas as pd
import numpy as np

# Create a minimal DataFrame that mimics the structure expected by build_graph_covid
# The function uses .iloc with indexing and accesses 'month' and 'year' columns
data_dict = {
    'month': ['January', 'February', 'March', 'April', 'May', 'June', 'July', 'August', 'September', 'October'],
    'year': [2020, 2020, 2020, 2020, 2020, 2020, 2020, 2020, 2021, 2021]
}
original_data = pd.DataFrame(data_dict)

# Create a raw_time_series with some values (function finds first non-zero/non-falsy value)
# The function uses enumerate and checks if x is truthy, then finds max with operator.itemgetter(1)
raw_time_series = [0, 10, 25, 50, 100, 75, 30, 15, 5, 2]

# wave_data: list of [start_index, end_index] tuples
# The function processes pairs (idx % 2 == 0 for starts)
wave_data = [
    [1, 3],
    [4, 6]
]

# trend_data: list of [start_index, end_index, trend_type] tuples
# Similarly processed in pairs
trend_data = [
    [1, 3, 'INCREASING'],
    [4, 6, 'DECREASING']
]

# content: a string label for the data type (e.g., 'COVID Cases')
content = 'COVID Cases'

# location: a string for the geographic location
location = 'New York'

ARGS = {
    'content': content,
    'location': location,
    'wave_data': wave_data,
    'trend_data': trend_data,
    'original_data': original_data,
    'raw_time_series': raw_time_series
}
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 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
build_graph_covid_form1 Ran
ran on a synthesized input
Mandar-Sharma/TCube/data2graph.py
code served (permissive licence) · get_code("92b8b52b495f227a")
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 numpy as np
import pandas as pd

# Create a minimal DataFrame that mimics the structure expected by the function
# The function expects original_data to have 'month' and 'year' columns
dates = pd.DataFrame({
    'month': ['January', 'February', 'March', 'April', 'May', 'June', 
              'July', 'August', 'September', 'October', 'November', 'December'] * 2,
    'year': [2020] * 12 + [2021] * 12
})

# Create time series data
raw_time_series = np.array([0, 5, 10, 15, 20, 25, 30, 28, 26, 24, 22, 20,
                            18, 16, 14, 12, 10, 8, 6, 4, 2, 1, 0, 0], dtype=float)

# Create wave data - list of [start_index, end_index] pairs
wave_data = [
    [0, 8],
    [8, 16],
    [16, 24]
]

# Create trend data - list of [start_index, end_index, trend_type] tuples
trend_data = [
    [0, 4, 'increasing'],
    [4, 8, 'decreasing'],
    [8, 12, 'decreasing'],
    [12, 24, 'decreasing']
]

ARGS = {
    'content': 'COVID cases',
    'location': 'New York',
    'wave_data': wave_data,
    'trend_data': trend_data,
    'original_data': dates,
    'raw_time_series': raw_time_series
}
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 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
build_graph_covid_nums Ran
ran on a synthesized input
Mandar-Sharma/TCube/data2graph.py
code served (permissive licence) · get_code("6abc9b7a7ae68831")
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 pandas as pd
import numpy as np

# Create a small pandas DataFrame to simulate original_data
# The function expects 'month' and 'year' columns
original_data = pd.DataFrame({
    'month': ['January', 'February', 'March', 'April', 'May', 'June', 'July', 'August'],
    'year': [2020, 2020, 2020, 2020, 2020, 2021, 2021, 2021]
})

# Create raw_time_series: a list of numbers representing case counts
# First element is 0 (no cases), then increasing values
raw_time_series = [0, 10, 25, 50, 100, 80, 40, 20]

# Create wave_data: list of [start_index, end_index] pairs for waves
# Each pair represents the start and end of a wave
wave_data = [
    [0, 4],  # First wave
    [4, 8]   # Second wave
]

# Create trend_data: list of [start_index, end_index, trend_type] for trends
# trend_type should be a string like 'increasing' or 'decreasing'
trend_data = [
    [0, 4, 'increasing'],
    [4, 8, 'decreasing']
]

# content is a string describing what is being measured
content = 'COVID-19 Cases'

# location is a string for the location
location = 'New York'

ARGS = {
    'content': content,
    'location': location,
    'wave_data': wave_data,
    'trend_data': trend_data,
    'original_data': original_data,
    'raw_time_series': raw_time_series
}
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 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
continuous_clusters Ran
ran on a synthesized input
Mandar-Sharma/TCube/utils.py
code served (permissive licence) · get_code("c64807c54898b8c7")
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 int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function expects segments to be an iterable of segments,
# where each segment is an iterable of "pieces" (could be any hashable object)
# The function checks len(segment) > 10 to decide behavior

# Create segments with some having length > 10 and some with length <= 10
segment1 = list(range(15))  # length 15, > 10
segment2 = list(range(100, 105))  # length 5, <= 10
segment3 = list(range(200, 212))  # length 12, > 10
segment4 = list(range(300, 303))  # length 3, <= 10

segments = [segment1, segment2, segment3, segment4]

ARGS = {
    "segments": segments
}
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 int — 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
find_contiguous Ran
ran on a synthesized input
Mandar-Sharma/TCube/utils.py
code served (permissive licence) · get_code("83d80ced1b946185")
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

class MockKMeans:
    def __init__(self, labels_list):
        self.labels_ = labels_list

# Create a mock kmeans object with labels that will produce segments
# We need labels that map to colors in the associated_colors dict
# Let's create a pattern: [0, 0, 1, 1, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1]
# This will create segments where some are long (>10) and some are short
kmeans_labels = MockKMeans([0, 0, 1, 1, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1])

ARGS = {
    "kmeans_labels": kmeans_labels
}
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
find_contiguous_sporadic Ran
ran on a synthesized input
Mandar-Sharma/TCube/utils.py
code served (permissive licence) · get_code("b0dfc4d5b264b383")
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
from sklearn.cluster import KMeans

torch.manual_seed(0)

# Create a mock kmeans_labels object with a labels_ attribute
class MockKMeansLabels:
    def __init__(self, labels):
        self.labels_ = labels

# Create sample labels for kmeans
# Using a mix of cluster assignments (0-13) that will map to colors
sample_labels = np.array([0, 0, 1, 1, 2, 2, 0, 0, 3, 3, 1, 1])

kmeans_labels = MockKMeansLabels(sample_labels)

ARGS = {
    "kmeans_labels": kmeans_labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, 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
randomise_data Ran
ran on a synthesized input
Yale-LILY/dart/evaluation/webnlg-automatic-evaluation/significance_block_creation.py
code served (permissive licence) · get_code("228074967276fcb2")
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:
# - filelines: a list that gets zipped with one of the index_shuf_* lists
#   The function sorts by the index list and extracts the filelines in that order
#   So filelines needs to be a list with enough elements to match the index list
# - param: a string that determines which index_shuf_* list to use

# For 'all-cat': needs 1862 elements (range(1, 1863))
# For 'old-cat': needs 971 elements (range(1, 972))
# For 'new-cat': needs 891 elements (range(1, 892))

# We'll create a test with the smallest case to keep things simple
filelines_all = [f"line_{i}" for i in range(1862)]
filelines_old = [f"line_{i}" for i in range(971)]
filelines_new = [f"line_{i}" for i in range(891)]

param_value = 'all-cat'

ARGS = {
    'filelines': filelines_all,
    'param': param_value
}
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
randomise_meteor_ter_data Ran
ran on a synthesized input
Yale-LILY/dart/evaluation/webnlg-automatic-evaluation/significance_block_creation.py
code served (permissive licence) · get_code("d577fd3c5c587731")
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

# Analyze the function:
# - filelines: a list that gets chunked into groups of 3
# - param: a string that should be one of 'all-cat', 'old-cat', or 'new-cat'
# The function chunks filelines into groups of 3, then reorders them based on shuffled indices

# Create a minimal filelines list that divides evenly by 3
# Using 'all-cat' which requires len(filelines) to work with index_shuf_all (range 1-1863, so 1862 items)
# Let's use a small multiple of 3 for testing
filelines = [f"line_{i}" for i in range(9)]  # 9 lines = 3 groups of 3

# Create a minimal param string
param = 'all-cat'

ARGS = {
    'filelines': filelines,
    'param': param
}
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
sumsquared_error Ran
ran on a synthesized input
Mandar-Sharma/TCube/segmentation.py
code served (permissive licence) · get_code("b6e7bc1c6a5b601a")
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

# Construct arguments for sumsquared_error(sequence, segment)
# sequence: a sequence (array-like) of numeric values
# segment: a tuple of (x0, y0, x1, y1) where x0 and x1 define the range

# Create a simple sequence with some linear trend
sequence = np.array([1.0, 2.1, 3.2, 4.0, 5.1, 6.2, 7.0, 8.1], dtype=float)

# Define a segment with x0=0, x1=4 (so we fit line to indices 0-4)
# y0 and y1 are not actually used by the function, so we set them arbitrarily
segment = (0, 0.0, 4, 5.0)

ARGS = {
    "sequence": sequence,
    "segment": segment,
}
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
generate Not yet run
unverified (harvested, no recorded run)
Mandar-Sharma/TCube/finetuned.py
code served (permissive licence) · get_code("96907642b314c1ca")
How it ran

no invocation recorded.

generate_ngram_es Not yet run
unverified (harvested, no recorded run)
Mandar-Sharma/TCube/finetuned.py
code served (permissive licence) · get_code("0e500eee8b2dc3c6")
How it ran

no invocation recorded.

generate_topk Not yet run
unverified (harvested, no recorded run)
Mandar-Sharma/TCube/finetuned.py
code served (permissive licence) · get_code("5cceb99692bc3aba")
How it ran

no invocation recorded.

leastsquareslinefit Not yet run
unverified (harvested, no recorded run)
Mandar-Sharma/TCube/segmentation.py
code served (permissive licence) · get_code("b6bbd4d2a80df557")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 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 14 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

We present DART, an open domain structured DAta Record to Text generation dataset with over 82k instances (DARTs). Data-to-Text annotations can be a costly process, especially when dealing with tables which are the major source of structured data and contain nontrivial structures. To this end, we propose a procedure of extracting semantic triples from tables that encodes their structures by exploiting the semantic dependencies among table headers and the table title. Our dataset construction framework effectively merged heterogeneous sources from open domain semantic parsing and dialogue-act-based meaning representation tasks by utilizing techniques such as: tree ontology annotation, question-answer pair to declarative sentence conversion, and predicate unification, all with minimum post-editing. We present systematic evaluation on DART as well as new state-of-the-art results on WebNLG 2017 to show that DART (1) poses new challenges to existing data-to-text datasets and (2) facilitates out-of-domain generalization. Our data and code can be found at https://github.com/Yale-LILY/dart.

For agents

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

get_harvested_code_for_paper("2007.02871")
get_code_for_paper("2007.02871")
have("2007.02871")

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

curl https://syntology.ai/api/ran/2007.02871.json

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

[![Syntology run record](https://syntology.ai/api/ran/2007.02871.svg)](https://syntology.ai/paper/2007.02871)

Connect an agent — have() is free.