SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.05935 · NAACL · 2025

DETQUS: Decomposition-Enhanced Transformers for QUery-focused Summarization

Rohan Sharma, Bonnie Dorr, Yasir Khan, Xinlei Wu, Sangpil Youm, Justin Ho, Aryaan Shaikh, Jairo Garciga

arXiv · PDF · Open in the Atlas

Code that ran

19 of 29 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 19 have no contract verdict; 0 failures of our instrument; 10 no recorded run), measured 2026-11-22; 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/2503.05935.json · badge: /api/ran/2503.05935.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
Rassibassi/claude canonical (classifier label)
classifier basis not recorded on the link
13 ran · 4 no recorded run
mrirecon/bart canonical (classifier label)
classifier basis not recorded on the link
6 ran · 6 no recorded run
FunctionStatusWhere it lives
PathCorrection Ran
ran on a synthesized input
mrirecon/bart/python/wslsupport.py
code served (permissive licence) · get_code("4ba47e8a93984c31")
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 string
import os

def PathCorrection(inData):
    outData = inData
    for i in string.ascii_lowercase:
        outData = outData.replace(i + ':', '/mnt/' + i)
        outData = outData.replace(i.upper() + ':', '/mnt/' + i)
    outData = outData.replace(os.path.sep, '/')
    return outData

ARGS = {
    "inData": "C:\\Users\\test\\data\\file.txt"
}
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 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
alphalin Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/models/SimpleXpmModel.py
code served (permissive licence) · get_code("950227152fdf8054")
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

def alphalin(alpha):
    return alpha / (10 * np.log10(np.exp(1))) * 0.001

# The function takes a single parameter 'alpha' and performs numerical operations on it
# alpha should be a numeric value (int, float, or numpy scalar)

ARGS = {
    "alpha": 0.5
}
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
analyse_d_string Ran
ran on a synthesized input
mrirecon/bart/python/splines_from_svg.py
code served (permissive licence) · get_code("972958eb73b4ee25")
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.

# Analyze the function to understand what it expects
# analyse_d_string takes a d_string parameter which is a string
# The function parses SVG path 'd' attribute strings
# It expects space-separated content with commands and coordinate pairs

# Create a valid SVG path 'd' string with various path commands
d_string = "M 10,20 L 30,40 C 50,60 70,80 90,100 h 10 v 20 L 150,200"

ARGS = {
    "d_string": d_string
}
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 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
bart2 Ran
ran on a synthesized input
mrirecon/bart/python/bart.py
code served (permissive licence) · get_code("4f1e246b50de93aa")
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 coroutine. 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 sys
from pathlib import Path
import tempfile as tmp

# Mock the cfl module since it's imported in the function
class MockCFL:
    @staticmethod
    def writecfl(filename, data):
        """Mock write - just create empty files"""
        if filename.endswith('.ra'):
            Path(filename).touch()
        else:
            Path(filename + '.hdr').touch()
            Path(filename + '.cfl').touch()
    
    @staticmethod
    def readcfl(filename):
        """Mock read - return dummy numpy array"""
        return np.zeros((2, 3, 4), dtype=np.complex64)

# Inject the mock cfl module
sys.modules['cfl'] = MockCFL()
cfl = MockCFL()

# Mock PathCorrection function
def PathCorrection(path):
    """Mock path correction for WSL"""
    return path

# Setup arguments for the bart function
ARGS = {
    'nargout': 1,  # Request 1 output
    'cmd': 'version',  # A simple BART command that doesn't require inputs
    # No additional positional args for version command
}
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 coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine 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
calcAseNoisePower Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/models/NLIN.py
code served (permissive licence) · get_code("8d584fc1fc015ba9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

class Param:
    def __init__(self, spanLength, alpha, nSpans, nPol, noiseFigure, lambda_, Rs):
        self.spanLength = spanLength
        self.alpha = alpha
        self.nSpans = nSpans
        self.nPol = nPol
        self.noiseFigure = noiseFigure
        self.lambda_ = lambda_
        self.Rs = Rs

param = Param(
    spanLength=80.0,
    alpha=0.2,
    nSpans=10,
    nPol=2,
    noiseFigure=5.5,
    lambda_=1.55e-6,
    Rs=32.0
)

ARGS = {
    "param": param
}
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 float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dispF Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/models/SimpleXpmModel.py
code served (permissive licence) · get_code("7cb293344cfcff60")
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 (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 numpy as np
import torch

torch.manual_seed(0)

# Analyzing the function:
# dispF takes three parameters: beta2, length, omega
# All are used in arithmetic operations with numpy
# beta2: scalar or array used in multiplication
# length: scalar or array used in multiplication
# omega: array used in exponentiation and multiplication

beta2 = 0.5
length = 2.0
omega = np.array([0.1, 0.2, 0.3, 0.4])

ARGS = {
    "beta2": beta2,
    "length": length,
    "omega": omega
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (4,) — shape (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
freqAxis Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/utils.py
code served (permissive licence) · get_code("9dc6794a711d46c3")
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 (16,) 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

# The function freqAxis takes two parameters:
# N: appears to be an integer representing the number of samples
# Fs: appears to be a sampling frequency (numeric value)

# Based on the function body:
# - N is used in arithmetic operations (N/2, modulo check, arange calls)
# - Fs is used in division (N/Fs)
# Both should be numeric values

ARGS = {
    "N": 16,      # Even integer for FFT-like frequency axis
    "Fs": 1000    # Sampling frequency in Hz
}
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 (16,) — shape (16,) 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
generateSymbols Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/tx.py
code served (permissive licence) · get_code("a639fc2fad8cf18d")
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

# The function expects:
# - constellation: a numpy array with shape (rows, cols) that supports indexing with [0, ...]
# - size: an integer specifying how many symbols to generate

# Looking at the function body:
# - constellation.shape[0] is accessed to get M
# - constellation[0, np.random.randint(M, size=size)] is used to index
# This means constellation should be 2D with first dimension > 1

# Create a small constellation array (complex numbers are typical for constellations)
constellation = np.array([
    [1.0 + 1.0j, 1.0 - 1.0j, -1.0 + 1.0j, -1.0 - 1.0j],  # row 0
    [0.5 + 0.5j, 0.5 - 0.5j, -0.5 + 0.5j, -0.5 - 0.5j]   # row 1
])

# Size parameter - number of symbols to generate
size = 3

ARGS = {
    "constellation": constellation,
    "size": size
}
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
i_fft Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/helper.py
code served (permissive licence) · get_code("3ec5d7f823add78d")
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 1, dtype family float, shape (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

# Create small complex-valued tensors for FFT operations
# h represents a filter, signal represents input signal
h = tf.constant(np.array([1.0 + 0.5j, 0.5 + 0.2j, 0.3 + 0.1j, 0.2 + 0.4j]), dtype=tf.complex64)
signal = tf.constant(np.array([0.5 + 0.1j, 0.3 + 0.2j, 0.8 + 0.3j, 0.2 + 0.5j]), dtype=tf.complex64)
transform_filter = True

ARGS = {
    "h": h,
    "signal": signal,
    "transform_filter": transform_filter,
}
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 1, dtype family float, shape (4,) — shape (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
logStepSizes Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/models/SSFstatic.py
code served (permissive licence) · get_code("a9616c656827ec50")
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 (5,) 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

# Analyze the function:
# - spanLength: used in arithmetic operations (float or int)
# - alpha: used in arithmetic operations (float or int)
# - nSteps: used in np.arange and arithmetic operations (int)

ARGS = {
    "spanLength": 10.0,
    "alpha": 0.5,
    "nSteps": 5
}
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 (5,) — shape (5,) 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
omegaAxis Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/utils.py
code served (permissive licence) · get_code("8212e21e3199c9d6")
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 (16,) 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

# omegaAxis takes two parameters: N and Fs
# N appears to be the number of samples (used in array construction and division)
# Fs appears to be the sampling frequency (used in division)

N = 16  # Even number for FFT-like use case
Fs = 1000.0  # Sampling frequency in Hz

ARGS = {
    "N": N,
    "Fs": Fs
}
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 (16,) — shape (16,) 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
qammod Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/tx.py
code served (permissive licence) · get_code("4a96f70bc450e463")
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 (1, 16) 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

# The function qammod takes a single parameter M
# M is used as: np.sqrt(M) - so it should be a number
# The function generates a QAM constellation with M points
# M should be a perfect square for standard QAM (e.g., 4, 16, 64, 256)

M = 16  # Valid QAM modulation order (16-QAM)

ARGS = {
    "M": M
}
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 (1, 16) — shape (1, 16) 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
randomizeSteps Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/models/SSFstatic.py
code served (permissive licence) · get_code("8c65a9541bded344")
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 numpy as np
import numpy.matlib as matlib

# Analyze the function:
# - stepSizes: used with matlib.repmat, indexing [0, 0], arithmetic operations
#   Should be a 2D array (1 row, multiple columns)
# - spanLength: scalar, used in subtraction
# - nSpans: scalar, used in repmat
# - sigma: scalar with default 0.01, used in multiplication

# Create small test arguments
stepSizes = np.array([[1.0, 2.0, 3.0]])  # 1x3 array of step sizes
spanLength = 6.0  # total span length that should be maintained
nSpans = 4  # number of spans to replicate
sigma = 0.01  # standard deviation for randomization

ARGS = {
    "stepSizes": stepSizes,
    "spanLength": spanLength,
    "nSpans": nSpans,
    "sigma": sigma
}
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 (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
readcfl Ran
ran on a synthesized input
mrirecon/bart/python/cfl.py
code served (permissive licence) · get_code("6f976527765f359e")
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 (4, 3, 2) 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 tempfile
import os

# Create temporary directory and files for testing
tmpdir = tempfile.mkdtemp()

# Create a .cfl/.hdr file pair
cfl_name = os.path.join(tmpdir, 'test')
hdr_path = cfl_name + '.hdr'
cfl_path = cfl_name + '.cfl'

# Write header file
dims = [4, 3, 2]
with open(hdr_path, 'wt') as f:
    f.write('# Dimensions\n')
    f.write(' '.join(str(d) for d in dims) + '\n')

# Write CFL file with complex64 data
data = np.arange(4*3*2, dtype=np.complex64)
data_reshaped = data.reshape(dims, order='F')
with open(cfl_path, 'wb') as f:
    data_reshaped.reshape(-1, order='F').tofile(f)

ARGS = {
    'name': cfl_name
}
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 3, dtype family float, shape (4, 3, 2) — shape (4, 3, 2) 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
rrcos Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/tx.py
code served (permissive licence) · get_code("289cb6311ecb08c5")
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 (41,) 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

ARGS = {
    "beta": 0.5,
    "sps": 4,
    "span": 10
}
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 (41,) — shape (41,) 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
tf1_op_exists Ran
ran on a synthesized input
mrirecon/bart/python/bart_tf.py
code served (permissive licence) · get_code("4c951d9cbee5929c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf

# Create a simple TensorFlow 1.x style graph
graph = tf.Graph()

# Add some operations to the graph
with graph.as_default():
    # Create a simple operation with a known name
    a = tf.constant(1.0, name="operation_1")
    b = tf.constant(2.0, name="operation_2")
    c = tf.add(a, b, name="add_op")

# The name parameter should be a string that represents an operation name in the graph
name = "add_op"

ARGS = {
    "graph": graph,
    "name": name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
try_float Ran
ran on a synthesized input
mrirecon/bart/python/splines_from_svg.py
code served (permissive licence) · get_code("b62bd82b6e59ae24")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function try_float takes a single parameter 'string' and attempts to convert it to float
# It returns True if successful, False if ValueError is raised

ARGS = {
    "string": "3.14"
}
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 bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
unitlessAxis Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/utils.py
code served (permissive licence) · get_code("d23692641fed1d6d")
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 (40,) 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

# Simple test setup for unitlessAxis function
# The function takes two numeric parameters: samplesPerSymbol and filterSpan

ARGS = {
    "samplesPerSymbol": 8,
    "filterSpan": 5
}
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 (40,) — shape (40,) 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
upsample Ran
ran on a synthesized input
Rassibassi/claude/claude/claude/claudeflow/helper.py
code served (permissive licence) · get_code("1f3baf4004725896")
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 3, dtype family float, shape (2, 3, 12) 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

# Create a small tensor with shape (2, 3, 4)
# The innermost dimension (4) represents N symbols that will be upsampled
symbols = tf.constant(np.random.randn(2, 3, 4).astype(np.float32))

# Samples per symbol (upsampling factor)
sps = 3

# Number of symbols
N = 4

ARGS = {
    "symbols": symbols,
    "sps": sps,
    "N": N,
}
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 3, dtype family float, shape (2, 3, 12) — shape (2, 3, 12) 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
QAMencoder Not yet run
unverified (harvested, no recorded run)
Rassibassi/claude/claude/claude/claudeflow/helper.py
code served (permissive licence) · get_code("c62786e7b48af2d7")
How it ran

no invocation recorded.

accumulatedOptimizer Not yet run
unverified (harvested, no recorded run)
Rassibassi/claude/claude/claude/claudeflow/training.py
code served (permissive licence) · get_code("3dc7a883d6ea54a0")
How it ran

no invocation recorded.

bart Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/bart.py
code served (permissive licence) · get_code("1e72cbe6b4e471ab")
How it ran

no invocation recorded.

bart_prepare Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/bart.py
code served (permissive licence) · get_code("f8a838574099ac03")
How it ran

no invocation recorded.

create_mean_metrics Not yet run
unverified (harvested, no recorded run)
Rassibassi/claude/claude/claude/claudeflow/training.py
code served (permissive licence) · get_code("e5fcb1eae7ed2d06")
How it ran

no invocation recorded.

create_reset_metric Not yet run
unverified (harvested, no recorded run)
Rassibassi/claude/claude/claude/claudeflow/training.py
code served (permissive licence) · get_code("19c5fa8d7e719d4a")
How it ran

no invocation recorded.

read_svg Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/splines_from_svg.py
code served (permissive licence) · get_code("e9e1487c469b3fb2")
How it ran

no invocation recorded.

readmulticfl Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/cfl.py
code served (permissive licence) · get_code("0a73ac0fca75f3fb")
How it ran

no invocation recorded.

tf1_find_tensors Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/bart_tf.py
code served (permissive licence) · get_code("2433e56a12a49752")
How it ran

no invocation recorded.

writecfl Not yet run
unverified (harvested, no recorded run)
mrirecon/bart/python/cfl.py
code served (permissive licence) · get_code("45651b708e62c1bf")
How it ran

no invocation recorded.

Also present in other papers' code

Of 29 samples lifted from this paper's linked repositories, 29 carry a function identity and 0 do not. Of the 29 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 29 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

Query-focused tabular summarization is an emerging task in table-to-text generation that synthesizes a summary response from tabular data based on user queries. Traditional transformer-based approaches face challenges due to token limitations and the complexity of reasoning over large tables. To address these challenges, we introduce DETQUS (Decomposition-Enhanced Transformers for QUery-focused Summarization), a system designed to improve summarization accuracy by leveraging tabular decomposition alongside a fine-tuned encoder-decoder model. DETQUS employs a large language model to selectively reduce table size, retaining only query-relevant columns while preserving essential information. This strategy enables more efficient processing of large tables and enhances summary quality. Our approach, equipped with table-based QA model Omnitab, achieves a ROUGE-L score of 0.4437, outperforming the previous state-ofthe-art REFACTOR model (ROUGE-L: 0.422). These results highlight DETQUS as a scalable and effective solution for query-focused tabular summarization, offering a structured alternative to more complex architectures.

For agents

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

get_harvested_code_for_paper("2503.05935")
get_code_for_paper("2503.05935")
have("2503.05935")

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

curl https://syntology.ai/api/ran/2503.05935.json

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

[![Syntology run record](https://syntology.ai/api/ran/2503.05935.svg)](https://syntology.ai/paper/2503.05935)

Connect an agent — have() is free.