|
|
--- |
|
|
license: mit |
|
|
size_categories: |
|
|
- n<1K |
|
|
task_categories: |
|
|
- question-answering |
|
|
- text-generation |
|
|
tags: |
|
|
- logical-reasoning |
|
|
- llm |
|
|
--- |
|
|
|
|
|
# SATQuest Dataset |
|
|
|
|
|
Paper: [SATQuest: A Verifier for Logical Reasoning Evaluation and Reinforcement Fine-Tuning of LLMs](https://huggingface.co/papers/2509.00930) |
|
|
|
|
|
[](https://opensource.org/licenses/MIT) |
|
|
[](https://github.com/sdpkjc/SATQuest) |
|
|
[](https://pypi.org/project/satquest/) |
|
|
|
|
|
**TL;DR.** Synthetic CNF benchmark for LLM reasoning: **140** matched SAT/UNSAT pairs with `n in [3, 16]` and fixed ratio `m=4n`. The dataset stores only CNF formulas and solver stats; use the [`SATQuest`](https://github.com/sdpkjc/SATQuest) Python library to render prompts/answers for SATDP, SATSP, MaxSAT, MCS, and MUS in four formats (math, DIMACS, story, dual story). |
|
|
|
|
|
## Data fields |
|
|
|
|
|
- `id`: unique identifier for each row. |
|
|
- `num_literal`: total number of literals in the unsatisfiable formula. |
|
|
- `sat_dimacs`: DIMACS representation of the satisfiable CNF. |
|
|
- `unsat_dimacs`: DIMACS representation of the unsatisfiable CNF. |
|
|
- `num_variable`: number of variables n, between 3 and 16. |
|
|
- `num_clause`: number of clauses `m` in the CNF, with `m=4n`. |
|
|
- `solver_metadatas`: dictionary of PySAT solver statistics (conflicts, decisions, propagations, restarts) for SATSP, MaxSAT, MCS, MUS, SATDP_SAT and SATDP_UNSAT tasks. |
|
|
|
|
|
## Quick usage |
|
|
|
|
|
Install the package and load a CNF instance: |
|
|
|
|
|
```python |
|
|
from datasets import load_dataset |
|
|
from satquest import CNF, create_problem, create_question |
|
|
|
|
|
item = load_dataset('sdpkjc/SATQuest', split='test')[0] |
|
|
cnf = CNF(dimacs=item['sat_dimacs']) |
|
|
|
|
|
problem = create_problem('SATSP', cnf) # or 'SATDP', 'MaxSAT', 'MCS', 'MUS' |
|
|
question = create_question('math') # or 'dimacs', 'story', 'dualstory' |
|
|
prompt = problem.accept(question) |
|
|
answer = problem.solution # reference answer |
|
|
reward = int(problem.check(answer)) # 1 if answer is correct, 0 otherwise |
|
|
``` |
|
|
|
|
|
## Reproducing the dataset |
|
|
|
|
|
To build an identical dataset from scratch, clone the repository and run the provided generation script: |
|
|
|
|
|
```bash |
|
|
pip install datasets numpy tyro |
|
|
git clone https://github.com/sdpkjc/SATQuest.git |
|
|
cd SATQuest |
|
|
|
|
|
# produce the evaluation set (140 CNF pairs) |
|
|
python gen_cnf_dataset.py --hf-entity {YOUR_HF_ENTITY} --seed 9527 |
|
|
``` |
|
|
|
|
|
## Citation |
|
|
|
|
|
Please cite the SATQuest paper if you use this dataset: |
|
|
|
|
|
```bibtex |
|
|
@misc{satquest2025, |
|
|
author = {Yanxiao Zhao, Yaqian Li, Zihao Bo, Rinyoichi Takezoe, Haojia Hui, Mo Guang, Lei Ren, Xiaolin Qin, Kaiwen Long}, |
|
|
title = {SATQuest: A Verifier for Logical Reasoning Evaluation and Reinforcement Fine-Tuning of LLMs}, |
|
|
year = {2025}, |
|
|
publisher = {Hugging Face}, |
|
|
journal = {Hugging Face Papers}, |
|
|
howpublished = {\url{https://huggingface.co/papers/2509.00930}}, |
|
|
} |
|
|
``` |
