metadata

language:
  - en
license: cc-by-sa-4.0
pretty_name: UQ
task_categories:
  - question-answering
tags:
  - stack-exchange
  - llm-evaluation
  - benchmark
  - reasoning
  - factuality

Dataset Card for UQ

Table of Contents
Dataset Description
- Dataset Summary
Dataset Structure
Sample Usage
Dataset Creation
Additional Information

Dataset Description

Repository: https://github.com/uq-project/UQ
Website: https://uq.stanford.edu
Paper: https://arxiv.org/abs/2508.17580

Dataset Summary

UQ is a collection of 500 popular unsolved questions from StackExchange sites covering Science, Technology, Culture & Recreation, Life & Arts, and Business.

Dataset Structure

Data Instances

An example looks as follows:

{
  'question_id': '5',
  'site': 'crypto',
  'title': 'Finding $x$ such that $g^x\\bmod p&lt;p/k$?',
  'body': "In a Schnorr group as used for DSA, of prime modulus $p$, prime order $q$, generator $g$ (with $p/g$ small), how can we efficiently exhibit an $x$ with $0<x<q$ such that $g^x\\bmod p<p/k$, for sizable $k$ but $k\\ll\\sqrt q$ ?

I see that for small $k$, it is enough to try incremental values of $x$ until finding an $x$ that fits, with expected cost $O(k)$ modular multiplications in $\\mathbb Z_p^*$. Is there better?

Assume $p$ is an $l$-bit prime; $q$ is an $n$-bit prime with $q$ dividing $p-1$; and $g=h^{(p-1)/q}$ for some (random?) $h$ in $\\mathbb Z_p^*$. For concrete values, assume $l\\approx1024$, $n\\approx160$, $k\\approx2^{64}$.

Update: I'm expecting the problem to be hard, and thus that the more difficult problem of exhibiting $x$ with $g^x\\bmod p$ in a range of width $p/k$ must be hard.\
\",
  'link': 'https://crypto.stackexchange.com/questions/48503/finding-x-such-that-gx-bmod-pp-k',
  'tags': ['discrete-logarithm', 'group-theory'],
  'votes': 19,
  'creation_date': datetime.datetime(2017, 6, 20, 8, 57, 11),
  'comments': ['Can you please provide me with some real values of variables in question instead of me attempting with random values.', '@pintor: no, I have made no progress nor got more feedback than the above comments.', "@fgrieu, did you get any more updates on the problem? I'm also curious", "This seems to reduce to a discrete logarithm problem in the multi-target setting, with a potentially very high target number depending on $k$. For a $k\\approx \\sqrt{p}$ index calculus seems like the best choice. Once you have one solution, getting more is relatively fast. Unsure if there's something better.", '@SamuelNeves: yes; the only constraint on $g^x\\bmod p$ is that it is small (about $\\log_2(k)$-bit smaller than $p$ is).', 'Do you get to choose what the $g^x \\mod p$ value is?', '@puzzlepalace: thank you for finding a hole; plugged it.', 'Choose $x = 0$ ? Runs in $\\Theta(1)$  ;)'],
  'comment_count': 8,
  'category': 'Technology',
  'diamond': 0,
}

Data Fields

The data fields are the same among all splits:

question_id: unique question id (1-500)
site: name of the StackExchange site
title: question title
body: question body
votes: StackExchange vote count of the question
creation_date: date of question creation on StackExchange
comments: StackExchange comments below the question
comment_count: Number of StackExchange comments
category: StackExchange category, one of {'Culture & Recreation', 'Science', 'Technology', 'Business', 'Life & Arts'}
diamond: Whether this question is included in UQ Diamond Subset

Data Splits

Category	Site	# Questions
Technology	Stack Overflow	21
	Mathematica	8
	Cryptography	5
	Retrocomputing	4
	Quantum Computing	4
	Space Exploration	3
	Unix & Linux	2
	TeX - LaTeX	2
	Code Golf	1
	Signal Processing	1
	Information Security	1
	Subtotal	52
Science	Math Overflow	200
	Mathematics	108
	Theoretical Computer Science	41
	Computer Science	12
	Cross Validated	9
	Physics	6
	Chemistry	4
	History of Science and Mathematics	3
	Linguistics	2
	Proof Assistants	2
	Artificial Intelligence	1
	Economics	1
	Bioacoustics	1
	Biology	1
	Medical Sciences	1
	Matter Modeling	1
	Operations Research	1
	Computational Science	1
	Subtotal	395
Culture & Recreation	Puzzling	8
	History	5
	Mythology & Folklore	2
	Role-playing Games	1
	Subtotal	16
Life & Arts	Science Fiction & Fantasy	35
Business	Quantitative Finance	2
Total	-	500

Sample Usage

You can load the data from 🤗 uq-project/uq via:

# pip install -q datasets
from datasets import load_dataset
dataset = load_dataset("uq-project/uq", split="test")

Additional Information

Licensing Information

The dataset is licensed under CC BY SA.

Citation Information

@misc{nie2025uqassessinglanguagemodels,
      title={UQ: Assessing Language Models on Unsolved Questions}, 
      author={Fan Nie and Ken Ziyu Liu and Zihao Wang and Rui Sun and Wei Liu and Weijia Shi and Huaxiu Yao and Linjun Zhang and Andrew Y. Ng and James Zou and Sanmi Koyejo and Yejin Choi and Percy Liang and Niklas Muennighoff},
      year={2025},
      eprint={2508.17580},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2508.17580}, 
}

Dataset Card for UQ

Table of Contents

Dataset Description

Dataset Summary

Dataset Structure

Data Instances

Data Fields

Data Splits

Sample Usage

Additional Information

Licensing Information

Citation Information