SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.08303 · NAACL · 2021

Capturing Row and Column Semantics in Transformer Based Question Answering over Tables

Saneem Chemmengath, Samarth Bharadwaj, Mustafa Canim, Michael Glass, Avirup Sil, Alfio Gliozzo, Vishwajeet Kumar, Nicolas Fauceglia, Rishav Chakravarti, Feifei Pan

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 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; 2 honoured the contract we drafted, 2 violated it and 0 have no contract verdict; 5 failures of our instrument; 1 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/2104.08303.json · badge: /api/ran/2104.08303.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
IBM/row-column-intersection canonical (classifier label)
classifier basis: census
9 ran · 1 no recorded run
FunctionStatusWhere it lives
avg Ran
ran, violated the contract we drafted
IBM/row-column-intersection/tableqa/aggregation_integration.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from ZhengyuZhao/TransferAttackEval, is pointer only (licence: NONE) · get_code("7a1a277b5455acbd")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from ZhengyuZhao/TransferAttackEval, 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.

convert_to_float Ran
ran, honoured the contract we drafted
IBM/row-column-intersection/tableqa/tapas_eval.py
code served (permissive licence) · get_code("67478c039bf8b9c0")
How it ran

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

import six

ARGS = {
    'value': '1,234.56'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
six — 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
get_numeric_col_vals Ran
ran; our contract draft was wrong, not the code
IBM/row-column-intersection/tableqa/aggregation_integration.py
code served (permissive licence) · get_code("3e6f5c4103652755")
How it ran

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

import numpy as np

# The function expects a table (list of rows, where each row is a list of strings)
# It will try to convert each cell to a number, and collect numeric column values

tbl = [
    ["10", "apple", "20.5"],
    ["15", "banana", "30,000"],
    ["20", "cherry", "40.0"]
]

ARGS = {
    "tbl": tbl
}
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 list — a list output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
log_softmax Ran
ran, violated the contract we drafted
IBM/row-column-intersection/tableqa/seq_pair_apply2tables.py
code served (permissive licence) · get_code("4b0deab5f5cb8d50")
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

# Construct a small numpy array as input
# The function expects a 1D or multi-dimensional array of numeric values
x = np.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    "x": x
}
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 (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
standard_json_mapper Ran
ran; our contract draft was wrong, not the code
IBM/row-column-intersection/dataloader/distloader_seq_pair.py
code served (permissive licence) · get_code("d994de88d7cad170")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 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

torch.manual_seed(0)

# The function takes a single parameter: jobj (a dictionary)
# It checks for 'text_b' key and returns different tuples based on that
# We need to provide a valid dictionary that matches one of the two cases

# Case 1: jobj with 'text_b' (returns 4-tuple)
jobj_with_text_b = {
    'id': 'doc_001',
    'text_a': 'This is text a',
    'text_b': 'This is text b',
    'label': 'positive'
}

ARGS = {
    'jobj': jobj_with_text_b
}
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, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
to_ndarray Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
IBM/row-column-intersection/tableqa/seq_pair_apply2tables.py
code served (permissive licence) · get_code("1be60c6154b0426b")
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

# Analyzing the function to_ndarray:
# - It takes 'preds' as a parameter
# - It expects preds to be iterable with elements that can be unpacked as (ndx, p)
# - It uses max() with key=lambda x: x[0], so each element must have at least 2 items
# - The first item (x[0]) is used as an index to find the max
# - It asserts len(preds) == num, where num = max_index + 1
# - It creates a numpy array of size num
# - It iterates through preds unpacking as (ndx, p) and assigns arr[ndx] = p

# Construct valid preds: list of (index, value) tuples
# If we have indices [0, 1, 2], then max index is 2, so num should be 3
# And len(preds) must equal 3

preds = [
    (0, 0.5),
    (1, 0.7),
    (2, 0.9)
]

ARGS = {
    "preds": preds
}
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
to_number Ran
ran, honoured the contract we drafted
IBM/row-column-intersection/tableqa/aggregation_integration.py
code served (permissive licence) · get_code("19b6aef837a2c77c")
How it ran

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

import torch
import numpy as np

def to_number(cell):
    return float(cell.replace(',', '').replace(' ', ''))

# The function takes a single parameter 'cell' which should be a string
# It removes commas and spaces, then converts to float
# We need to provide a string with commas and/or spaces

ARGS = {
    'cell': '1,234 567.89'
}
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 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
tokenize_string Ran
ran; our contract draft was wrong, not the code
IBM/row-column-intersection/tableqa/tapas_eval.py
code served (permissive licence) · get_code("84b82f831409eb64")
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 re

_TOKENIZER = re.compile('\\w+|[^\\w\\s]+', re.UNICODE)

def tokenize_string(x):
    return list(_TOKENIZER.findall(x.lower()))

ARGS = {
    'x': "Hello, World! This is a TEST 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 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
wtq_normalize Ran
ran; our contract draft was wrong, not the code
IBM/row-column-intersection/tableqa/tapas_eval.py
code served (permissive licence) · get_code("cd5b71c63186c124")
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 six
import re
import unicodedata

# The function wtq_normalize takes a single parameter x
# which can be an integer or string that gets normalized

ARGS = {
    "x": "WikiTableQuestions example text with [citation] and (parenthetical) content!"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
six — 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
sentence_to_inputs Not yet run
unverified (harvested, no recorded run)
IBM/row-column-intersection/dataloader/distloader_base.py
code served (permissive licence) · get_code("e8e58c66ec0109a2")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 2 later) and 7 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_to_float 2 other papers
1 dated earlier, 1 later
2004.02349 (2020-04)
TAPAS: Weakly Supervised Table Parsing via Pre-training
2306.14321 (2023-06)
ROBUT: A Systematic Study of Table QA Robustness Against Human-Annotated Adversarial Perturbations
avg 1 other paper
1 later
2310.17645 (2023-10)
PUBDEF: DEFENDING AGAINST TRANSFER ATTACKS FROM PUBLIC MODELS
tokenize_string 1 other paper
1 dated earlier
2004.02349 (2020-04)
TAPAS: Weakly Supervised Table Parsing via Pre-training

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

Transformer based architectures are recently used for the task of answering questions over tables. In order to improve the accuracy on this task, specialized pre-training techniques have been developed and applied on millions of open-domain web tables. In this paper, we propose two novel approaches demonstrating that one can achieve superior performance on table QA task without even using any of these specialized pre-training techniques. The first model, called RCI interaction, leverages a transformer based architecture that independently classifies rows and columns to identify relevant cells. While this model yields extremely high accuracy at finding cell values on recent benchmarks, a second model we propose, called RCI representation, provides a significant efficiency advantage for online QA systems over tables by materializing embeddings for existing tables. Experiments on recent benchmarks prove that the proposed methods can effectively locate cell values on tables (up to ∼98% Hit@1 accuracy on WikiSQL lookup questions). Also, the interaction model outperforms the state-of-the-art transformer based approaches, pre-trained on very large table corpora (TAPAS and TABERT), achieving ∼3.4% and ∼18.86% additional precision improvement on the standard WikiSQL benchmark 1 .

For agents

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

get_harvested_code_for_paper("2104.08303")
get_code_for_paper("2104.08303")
have("2104.08303")

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

curl https://syntology.ai/api/ran/2104.08303.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.08303.svg)](https://syntology.ai/paper/2104.08303)

Connect an agent — have() is free.