SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2010.04898 · 2020

Open-Domain Question Answering Goes Conversational via Question Rewriting

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 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 6 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2010.04898.json · badge: /api/ran/2010.04898.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
apple/ml-qrecc canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
chunk_doc Ran
ran on a synthesized input
apple/ml-qrecc/collection/paragraph_chunker.py
code served (permissive licence) · get_code("f445d680aeec8da8")
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

# Construct valid arguments for chunk_doc

# The function takes a single parameter: content (str)
# It splits the content by newlines, filters out lines with '===',
# skips empty lines, splits lines into tokens, and groups them into passages
# of at least MIN_PASSAGE_TOKENS (220) tokens each.

# Create a sample document with enough content to test the chunking logic
content = """
This is the first line of the document with some words.
Here is another line that adds more tokens to accumulate.
Yet another line to keep building up the token count.
We need many lines to reach the minimum passage token threshold.
The minimum is set to 220 tokens per passage so we need substantial content.
Let me add more lines with various words and phrases.
This document will test the chunking functionality properly.
More content here to ensure we have enough tokens.
Additional text to continue building the passage.
Even more words to reach the required token minimum.
Continuing with more substantial text content now.
Adding further lines with additional information here.
This keeps building up the total token count needed.
More and more content keeps accumulating properly.
We are getting closer to reaching 220 tokens threshold.
Additional text continues to be added systematically.
Each line contributes more tokens to the total count.
The chunking algorithm will split at appropriate places.
More lines ensure we test the full functionality properly.
Substantial content keeps being added to reach minimum.
The document now has sufficient tokens for proper testing.
Final lines ensure we have well over 220 tokens total.
This completes the content needed for the function test.
"""

ARGS = {
    "content": content
}
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
compute_exact Ran
ran on a synthesized input
apple/ml-qrecc/utils/evaluate_qa.py
code served (permissive licence) · get_code("0d6a04e3f9c96f42")
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 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
import string

def normalize_answer(s: str) -> str:
    """Lower text and remove punctuation, articles and extra whitespace."""

    def remove_articles(text):
        regex = re.compile('\\b(a|an|the)\\b', re.UNICODE)
        return re.sub(regex, ' ', text)

    def white_space_fix(text):
        return ' '.join(text.split())

    def remove_punc(text):
        exclude = set(string.punctuation)
        return ''.join((ch for ch in text if ch not in exclude))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_articles(remove_punc(lower(s))))

def compute_exact(a_gold: str, a_pred: str) -> int:
    """Compute the Exact Match score."""
    return int(normalize_answer(a_gold) == normalize_answer(a_pred))

ARGS = {
    'a_gold': 'The Quick Brown Fox',
    'a_pred': 'the quick brown fox'
}
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 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
compute_mean_reciprocal_rank Ran
ran on a synthesized input
apple/ml-qrecc/utils/evaluate_retrieval.py
code served (permissive licence) · get_code("0ebd0ec1f3ba37e6")
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

# Construct valid arguments for compute_mean_reciprocal_rank

# The function expects:
# - question_id_to_docs: Dict[str, List[dict]] where each dict has 'answer' (list) and 'f1' (float) and 'rank' (int/float)
# - relevance_threshold: float

question_id_to_docs = {
    "q1": [
        {"answer": ["yes"], "f1": 0.9, "rank": 1},
        {"answer": [], "f1": 0.5, "rank": 2},
        {"answer": ["no"], "f1": 0.3, "rank": 3},
    ],
    "q2": [
        {"answer": [], "f1": 0.7, "rank": 1},
        {"answer": ["correct"], "f1": 0.85, "rank": 2},
        {"answer": ["wrong"], "f1": 0.4, "rank": 3},
    ],
    "q3": [
        {"answer": ["maybe"], "f1": 0.2, "rank": 1},
        {"answer": ["sure"], "f1": 0.6, "rank": 2},
    ],
}

relevance_threshold = 0.5

ARGS = {
    "question_id_to_docs": question_id_to_docs,
    "relevance_threshold": relevance_threshold,
}
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 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
compute_recall_at_k Ran
ran on a synthesized input
apple/ml-qrecc/utils/evaluate_retrieval.py
code served (permissive licence) · get_code("2bc9d0de07492af0")
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

# Construct valid arguments for compute_recall_at_k

# The function expects:
# - question_id_to_docs: Dict[str, List[dict]]
#   Each dict in the list needs: 'answer' (string), 'f1' (float), 'rank' (int)
# - k: int (cutoff rank)
# - relevance_threshold: float (minimum f1 score to be considered relevant)

question_id_to_docs = {
    "q1": [
        {"answer": "Paris", "f1": 0.95, "rank": 1},
        {"answer": "", "f1": 0.5, "rank": 2},
        {"answer": "France", "f1": 0.85, "rank": 3},
    ],
    "q2": [
        {"answer": "", "f1": 0.3, "rank": 1},
        {"answer": "Berlin", "f1": 0.92, "rank": 2},
        {"answer": "Germany", "f1": 0.88, "rank": 3},
    ],
    "q3": [
        {"answer": "Tokyo", "f1": 0.6, "rank": 1},
        {"answer": "Japan", "f1": 0.7, "rank": 2},
    ],
}

k = 2
relevance_threshold = 0.8

ARGS = {
    "question_id_to_docs": question_id_to_docs,
    "k": k,
    "relevance_threshold": relevance_threshold,
}
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
get_tokens Ran
ran on a synthesized input
apple/ml-qrecc/utils/evaluate_qa.py
code served (permissive licence) · get_code("119c6dab40b4b162")
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
import string
from typing import List

def normalize_answer(s: str) -> str:
    """Lower text and remove punctuation, articles and extra whitespace."""

    def remove_articles(text):
        regex = re.compile('\\b(a|an|the)\\b', re.UNICODE)
        return re.sub(regex, ' ', text)

    def white_space_fix(text):
        return ' '.join(text.split())

    def remove_punc(text):
        exclude = set(string.punctuation)
        return ''.join((ch for ch in text if ch not in exclude))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_articles(remove_punc(lower(s))))

def get_tokens(s: str) -> List[str]:
    """Normalize string and split string into tokens."""
    if not s:
        return []
    return normalize_answer(s).split()

ARGS = {
    "s": "The quick brown fox jumps over the lazy dog!"
}
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
normalize_answer Ran
ran on a synthesized input
apple/ml-qrecc/utils/evaluate_qa.py
code served (permissive licence) · get_code("c6e13a9c4b963076")
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 re
import string

def normalize_answer(s: str) -> str:
    """Lower text and remove punctuation, articles and extra whitespace."""

    def remove_articles(text):
        regex = re.compile('\\b(a|an|the)\\b', re.UNICODE)
        return re.sub(regex, ' ', text)

    def white_space_fix(text):
        return ' '.join(text.split())

    def remove_punc(text):
        exclude = set(string.punctuation)
        return ''.join((ch for ch in text if ch not in exclude))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_articles(remove_punc(lower(s))))


ARGS = {
    "s": "The Quick Brown Fox! Is a famous pangram."
}
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
download_with_retry Not yet run
unverified (harvested, no recorded run)
apple/ml-qrecc/collection/download_wayback_passages.py
code served (permissive licence) · get_code("42b722ad97d9c139")
How it ran

no invocation recorded.

process_cc_index Not yet run
unverified (harvested, no recorded run)
apple/ml-qrecc/collection/download_commoncrawl_passages.py
code served (permissive licence) · get_code("6e2174fcdc2d0729")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 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 8 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 introduce a new dataset for Question Rewriting in Conversational Context (QReCC), which contains 14K conversations with 80K question-answer pairs. The task in QReCC is to find answers to conversational questions within a collection of 10M web pages (split into 54M passages). Answers to questions in the same conversation may be distributed across several web pages. QReCC provides annotations that allow us to train and evaluate individual subtasks of question rewriting, passage retrieval and reading comprehension required for the end-to-end conversational question answering (QA) task. We report the effectiveness of a strong baseline approach that combines the state-of-the-art model for question rewriting, and competitive models for open-domain QA. Our results set the first baseline for the QReCC dataset with F1 of 19.10, compared to the human upper bound of 75.45, indicating the difficulty of the setup and a large room for improvement.

For agents

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

get_harvested_code_for_paper("2010.04898")
get_code_for_paper("2010.04898")
have("2010.04898")

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

curl https://syntology.ai/api/ran/2010.04898.json

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

[![Syntology run record](https://syntology.ai/api/ran/2010.04898.svg)](https://syntology.ai/paper/2010.04898)

Connect an agent — have() is free.