A Dataset of Information-Seeking Questions and Answers Anchored in Research Papers
Paper • 2105.03011 • Published • 1
Datasets:
Tasks:
Question Answering
Modalities:
Text
Formats:
json
Sub-tasks:
extractive-qa
Languages:
English
Size:
1K - 10K
ArXiv:
License:
messages listlengths 2 2 | meta dict |
|---|---|
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Answer the question using the provided scientific paper context. Use only the given context, and if the answer is not supported, reply with \"Unanswerable\".\n\nQuestion: How did they induce the CFG?\n\nPaper title: Joint learni... | {
"dataset": "qasper",
"split": "train",
"paper_id": "1601.00901",
"question_id": "74b0d3ee0cc9b0a3d9b264aba9901ff97048a897",
"annotation_id": "2ad8a8c4827f8617a9f38d243b56a2adb3518ba5",
"worker_id": "c1fbdd7a261021041f75fbe00a55b4c386ebbbb4",
"answer_type": "extractive",
"context_mode": "evidence",
"... |
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Answer the question using the provided scientific paper context. Use only the given context, and if the answer is not supported, reply with \"Unanswerable\".\n\nQuestion: How big is their dataset?\n\nPaper title: Joint learning ... | {
"dataset": "qasper",
"split": "train",
"paper_id": "1601.00901",
"question_id": "9eb5b336b3dcb7ab63f673ba9ab1818573cce6c3",
"annotation_id": "5cd39c20891ea9876f10018e07223969332a4dfb",
"worker_id": "c1fbdd7a261021041f75fbe00a55b4c386ebbbb4",
"answer_type": "extractive",
"context_mode": "full_text",
... |
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Answer the question using the provided scientific paper context. Use only the given context, and if the answer is not supported, reply with \"Unanswerable\".\n\nQuestion: What benchmark datasets they use?\n\nPaper title: Learnin... | {
"dataset": "qasper",
"split": "train",
"paper_id": "1601.01705",
"question_id": "5712a0b1e33484ebc6d71c70ae222109c08dede2",
"annotation_id": "46e5ea281159cbd3c2de92d958a1f2a835593916",
"worker_id": "057bf5a20e4406f1f05cf82ecd49cf4f227dd287",
"answer_type": "free_form",
"context_mode": "full_text",
"... |
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Answer the question using the provided scientific paper context. Use only the given context, and if the answer is not supported, reply with \"Unanswerable\".\n\nQuestion: What languages did they experiment with?\n\nPaper title: ... | {
"dataset": "qasper",
"split": "train",
"paper_id": "1601.02166",
"question_id": "e67d2266476abd157fc8c396b3dfb70cb343471e",
"annotation_id": "242af0098d4f61a8d54d7f9225e27a11d1652f95",
"worker_id": "5d0eb97e8e840e171f73b7642c2c89dd3984157b",
"answer_type": "unanswerable",
"context_mode": "evidence",
... |
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Answer the question using the provided scientific paper context. Use only the given context, and if the answer is not supported, reply with \"Unanswerable\".\n\nQuestion: what bottlenecks were identified?\n\nPaper title: Evaluat... | {
"dataset": "qasper",
"split": "train",
"paper_id": "1601.02543",
"question_id": "8d793bda51a53a4605c1c33e7fd20ba35581a518",
"annotation_id": "40e8b561276384d47c4e6a0d3861d502b1fe37f3",
"worker_id": "c1018a31c3272ce74964a3280069f62f314a1a58",
"answer_type": "free_form",
"context_mode": "evidence",
"e... |
[{"role":"user","content":[{"type":"text","text":"Answer the question using the provided scientific (...TRUNCATED) | {"dataset":"qasper","split":"train","paper_id":"1601.04012","question_id":"eac042734f76e787cb98ba3d0(...TRUNCATED) |
[{"role":"user","content":[{"type":"text","text":"Answer the question using the provided scientific (...TRUNCATED) | {"dataset":"qasper","split":"train","paper_id":"1601.06068","question_id":"117aa7811ed60e84d40cd8f9c(...TRUNCATED) |
[{"role":"user","content":[{"type":"text","text":"Answer the question using the provided scientific (...TRUNCATED) | {"dataset":"qasper","split":"train","paper_id":"1601.06068","question_id":"c359ab8ebef6f60c5a38f5244(...TRUNCATED) |
[{"role":"user","content":[{"type":"text","text":"Answer the question using the provided scientific (...TRUNCATED) | {"dataset":"qasper","split":"train","paper_id":"1601.06068","question_id":"ad362365656b0b218ba324ae6(...TRUNCATED) |
[{"role":"user","content":[{"type":"text","text":"Answer the question using the provided scientific (...TRUNCATED) | {"dataset":"qasper","split":"train","paper_id":"1601.06068","question_id":"423bb905e404e88a168e7e807(...TRUNCATED) |
End of preview. Expand in Data Studio
QASPER (Chat-Format Preparation)
This dataset is a chat-format preparation of QASPER for supervised fine-tuning (SFT).
Format
This format is commonly referred to as:
- chat-format SFT data
- instruction-tuning conversations
- OpenAI-style
messagesformat
Included files
train.jsonlvalidation.jsonlstats.jsonprepare_qasper_unsloth.py
Source
- Base dataset:
allenai/qasper
Original Dataset Highlights
- Original dataset:
allenai/qasper - Focus: question answering on scientific NLP papers with evidence selection.
- Reported scale on source card: 5,049 questions over 1,585 papers.
- Key annotation properties: multiple answer types (free-form, extractive, yes/no, unanswerable) and evidence annotations.
- Paper: A Dataset of Information-Seeking Questions and Answers Anchored in Research Papers
Preparation summary
- One row per
(paper, question)using the best available annotation. - Answer normalization priority:
- free-form
- yes/no
- extractive spans
- unanswerable
- Context mode is mixed between:
- evidence-only
- full-text
- User prompt follows a question-first structure.
Assistant target is the normalized answer text.
Schema
Each JSONL row contains:
messagesuser: text instruction + question + title + abstract + contextassistant: text answer
meta: ids, answer type, context mode, evidence count
Reproduction
python prepare_qasper_unsloth.py
- Downloads last month
- 26