SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.03490 · EMNLP · 2020

What do Models Learn from Question Answering Datasets?

Amir Saffari, Priyanka Sen

arXiv · PDF · Open in the Atlas

Code that ran

5 of 6 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, 1 violated it and 1 has no contract verdict; 2 failures of our instrument; 1 no recorded run), measured 2026-11-21; 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/2004.03490.json · badge: /api/ran/2004.03490.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
amazon-science/qa-dataset-converter canonical (classifier label)
classifier basis: classifier_2025
4 ran · 0 no recorded run
amazon-research/qa-dataset-converter canonical (classifier label)
classifier basis: census
1 ran · 1 no recorded run
FunctionStatusWhere it lives
answer_index_in_document Ran
ran on a synthesized input
amazon-research/qa-dataset-converter/triviaqa/triviaqa_to_squad.py
code served (permissive licence) · get_code("d3c04917ac91d32d")
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

ARGS = {
    'answer': {
        'Aliases': ['John Smith', 'J. Smith'],
        'NormalizedAliases': ['john smith', 'js'],
        'NormalizedValue': 'john smith'
    },
    'document': 'The person named john smith was born in 1990. John Smith is a common name.'
}
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
answer_start Ran
ran, honoured the contract we drafted
amazon-science/qa-dataset-converter/newsqa/newsqa_to_squad.py
code served (permissive licence) · get_code("20640d3656e293c0")
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) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def answer_text(story_text, answer_token_ranges):
    story_text_list = story_text.split()
    if answer_token_ranges:
        token_range = list(map(int, re.split(':|,', answer_token_ranges)))[0:2]
        answer = ' '.join(story_text_list[slice(token_range[0], token_range[1])])
        return answer

def answer_start(story_text, answer_text):
    if answer_text:
        return story_text.find(answer_text)
    else:
        return -1

# Setup code
story_text = "The quick brown fox jumps over the lazy dog"
answer_text_value = "brown fox"

ARGS = {
    "story_text": story_text,
    "answer_text": answer_text_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
answer_text Ran
ran; our contract draft was wrong, not the code
amazon-science/qa-dataset-converter/newsqa/newsqa_to_squad.py
code served (permissive licence) · get_code("d9a2ef2b7ba09e79")
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) 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 re

# Analyze the function:
# - story_text: a string that gets split() into a list of words
# - answer_token_ranges: a string that gets parsed with re.split(':|,', ...) 
#   and converted to ints, then sliced to get first 2 elements
#   These are used as slice indices into story_text_list

# Create valid arguments
story_text = "The quick brown fox jumps over the lazy dog and runs away"
answer_token_ranges = "2:5"

ARGS = {
    "story_text": story_text,
    "answer_token_ranges": answer_token_ranges
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
has_long_answer Ran
ran, violated the contract we drafted
amazon-science/qa-dataset-converter/nq/nq_to_squad.py
code served (permissive licence) · get_code("8c01993af488673e")
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) 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

# Construct a valid nq_example with a single annotation that has a long answer
nq_example_single = {
    'annotations': [
        {
            'long_answer': {
                'start_byte': 42
            }
        }
    ]
}

# Construct a valid nq_example with multiple annotations, at least 2 with long answers
nq_example_multiple = {
    'annotations': [
        {
            'long_answer': {
                'start_byte': 10
            }
        },
        {
            'long_answer': {
                'start_byte': 20
            }
        },
        {
            'long_answer': {
                'start_byte': -1
            }
        }
    ]
}

# Test with single annotation case
ARGS = {
    'nq_example': nq_example_single
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
has_short_answer Ran
ran; our contract draft was wrong, not the code
amazon-science/qa-dataset-converter/nq/nq_to_squad.py
code served (permissive licence) · get_code("a4f3efffaf7454f7")
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) 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

# Construct a valid nq_example with single annotation and short answers
nq_example_with_short_answers = {
    'annotations': [
        {
            'short_answers': ['Paris'],
            'yes_no_answer': 'NONE'
        }
    ]
}

# Construct a valid nq_example with single annotation and yes/no answer
nq_example_with_yes_no = {
    'annotations': [
        {
            'short_answers': [],
            'yes_no_answer': 'YES'
        }
    ]
}

# Construct a valid nq_example with multiple annotations where >= 2 have answers
nq_example_multiple_with_answers = {
    'annotations': [
        {
            'short_answers': ['Answer 1'],
            'yes_no_answer': 'NONE'
        },
        {
            'short_answers': [],
            'yes_no_answer': 'NO'
        },
        {
            'short_answers': [],
            'yes_no_answer': 'NONE'
        }
    ]
}

ARGS = {
    'nq_example': nq_example_with_short_answers
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
select_relevant_portion Not yet run
unverified (harvested, no recorded run)
amazon-research/qa-dataset-converter/triviaqa/triviaqa_to_squad.py
code served (permissive licence) · get_code("486151e6c8d08794")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 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 6 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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

While models have reached superhuman performance on popular question answering (QA) datasets such as SQuAD, they have yet to outperform humans on the task of question answering itself. In this paper, we investigate if models are learning reading comprehension from QA datasets by evaluating BERT-based models across five datasets. We evaluate models on their generalizability to out-of-domain examples, responses to missing or incorrect data, and ability to handle question variations. We find that no single dataset is robust to all of our experiments and identify shortcomings in both datasets and evaluation methods. Following our analysis, we make recommendations for building future QA datasets that better evaluate the task of question answering through reading comprehension. We also release code to convert QA datasets to a shared format for easier experimentation at https: //github.com/amazon-research/ qa-dataset-converter.

For agents

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

get_harvested_code_for_paper("2004.03490")
get_code_for_paper("2004.03490")
have("2004.03490")

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

curl https://syntology.ai/api/ran/2004.03490.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.03490.svg)](https://syntology.ai/paper/2004.03490)

Connect an agent — have() is free.