ReSyn Dataset

This repository contains the official datasets for the paper ReSyn: A Generalized Recursive Regular Expression Synthesis Framework (accepted at IJCAI 2026).

• Paper: ReSyn: A Generalized Recursive Regular Expression Synthesis Framework
• GitHub: mrseongminkim/ReSyn

Dataset Description

ReSyn is a dataset designed for Programming-By-Example (PBE) systems focused on regular expression synthesis. It aims to capture the high structural complexity of real-world regexes, such as deep nesting and frequent use of union operations, which are often missing from simplified benchmarks.

The framework uses a divide-and-conquer approach to decompose complex synthesis problems into manageable sub-problems by adaptively predicting whether to split examples sequentially (Concatenation) or group them by structural similarity (Union).

Splits

The dataset consists of the following splits:

• train: 1,410,280 examples used for training the framework's core components.
• valid: 32,236 examples for validation.
• benchmark: 2,728 real-world regex synthesis problems for evaluating performance on structurally complex patterns.

Each example includes a target regex, as well as sets of positive_strings and negative_strings.

Citation

If you use this dataset in your research, please cite:

@inproceedings{kim2026resyn,
  title={ReSyn: A Generalized Recursive Regular Expression Synthesis Framework},
  author={Kim, Seongmin and Cheon, Hyunjoon and Kim, Su-Hyeon and Han, Yo-Sub and Ko, Sang-Ki},
  booktitle={Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-26)},
  year={2026}
}