---
title: "DataArc-SynData-Toolkit — paper by Shi et al. (2026)"
description: "The authors present DataArc-SynData-Toolkit, an open-source system driven by configuration files. 20 references."
authors: ["Zhichao Shi", "Cehao Yang", "Hao Zhou", "Xiaojun Wu", "Huajie Li", "Xuhui Jiang", "Chengjin Xu", "Yuanzhuo Wang", "Jian Guo"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)"
source: "https://aclanthology.org/2026.acl-demo.31/"
doi: "10.18653/v1/2026.acl-demo.31"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/shi-2026-dataarc-syndata-toolkit-unified-closed/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/shi-2026-dataarc-syndata-toolkit-unified-closed/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/26/) / DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis

# DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis

**Authors:** [Zhichao Shi](https://md.research.elamur.ai/free/en/authors/zhichao-shi/), [Cehao Yang](https://md.research.elamur.ai/free/en/authors/cehao-yang/), [Hao Zhou](https://md.research.elamur.ai/free/en/authors/hao-zhou/), [Xiaojun Wu](https://md.research.elamur.ai/free/en/authors/xiaojun-wu/), Huajie Li, [Xuhui Jiang](https://md.research.elamur.ai/free/en/authors/xuhui-jiang/), [Chengjin Xu](https://md.research.elamur.ai/free/en/authors/chengjin-xu/), [Yuanzhuo Wang](https://md.research.elamur.ai/free/en/authors/yuanzhuo-wang/) and [Jian Guo](https://md.research.elamur.ai/free/en/authors/jian-guo/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-demo.31/) · [PDF](https://aclanthology.org/2026.acl-demo.31.pdf) · [DOI 10.18653/v1/2026.acl-demo.31](https://doi.org/10.18653/v1/2026.acl-demo.31) · [OpenAlex](https://openalex.org/W7166857999)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Synthetic data can ease data shortages for LLMs in specialised domains and low-resource languages, but current tools tend to be hard for non-experts, rely on inconsistent data formats, and handle extra modalities or languages poorly. The authors present DataArc-SynData-Toolkit, an open-source system driven by configuration files. In one loop it covers generating data, filtering and rewriting it for quality, post-training a model, and evaluating the result, and it can be run from a CLI or a web interface. Every generation route follows the same five-step abstraction: parse the task, prepare sources, build constraints, acquire data, and normalise the structure. The data can come from local corpora, Hugging Face datasets, or teacher-model distillation. Quality control scores each sample, rewrites it, and scores it again. In experiments, synthetic sets of 1,000–4,000 samples produced with gpt-5-mini raised scores on domain, multilingual and multimodal benchmarks. Gains held across model families and grew with data size. For teams with little data, such as in Ukrainian NLP, the toolkit offers a reusable and extensible way to bootstrap training sets.

## Key points
- **Generation paths and quality control:** three executors (LocalTaskExecutor, WebTaskExecutor, DistillTaskExecutor) inherit from BaseTaskExecutor. Quality control is extensible through the Evaluator and BaseRewriter classes, for example the difficulty-adjusting rewriter DifficultyAdjustRewriter.
- **Seed data and benchmarks:** seeds come from 5 candidate Hugging Face datasets, from which 25 instances are sampled. Benchmarks include MedQA, Flare-CFA, LexEval, ABB (Arabic) and FinMME.
- **Models and hardware:** Qwen2.5-7B-Instruct, Qwen3-4B, Qwen2.5-VL-7B-Instruct and Qwen3-VL-4B-Instruct, trained on eight H800 GPUs.
- **Main results with 4,000 samples:**
  - Qwen2.5-7B: MedQA rose from 42.34 to 68.12 and LexEval from 19.80 to 46.15.
  - Qwen3-4B: MedQA rose from 57.52 to 70.62 and LexEval from 31.48 to 48.00.
- **Speed:** without parallelism, ParallelExecutor needs 909 s for 500 samples. Running 10–25 workers cuts this time substantially.
- **Interface:** a FastAPI backend with a React frontend. Code is at github.com/DataArcTech/DataArc-SynData-Toolkit.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Institute of Computing Technology](https://md.research.elamur.ai/free/en/labs/institute-of-computing-technology/) — Research institute, Beijing, China · 51 works · 3,411 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works

🧭 Topics (2): [🌍 Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · 828 works · [🖼️ Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/) · 657 works

## Cite

Cite as: Zhichao Shi, Cehao Yang, Hao Zhou, Xiaojun Wu, Huajie Li, Xuhui Jiang, Chengjin Xu, Yuanzhuo Wang and Jian Guo. 2026. DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{shi2026dataarc,
  title = {{DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis}},
  author = {Zhichao Shi, Cehao Yang, Hao Zhou, Xiaojun Wu, Huajie Li, Xuhui Jiang, Chengjin Xu, Yuanzhuo Wang and Jian Guo},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-demo.31},
  url = {https://doi.org/10.18653/v1/2026.acl-demo.31}
}
```

- [← previous Data-Centric Approach at the LoResMT 2026 Turkic Translation Challenge: Russian-Kyrgyz](https://md.research.elamur.ai/free/en/papers/novokshanov-2026-data-centric-approach-loresmt-2026/)
- [next → DATASHI: A Parallel English–Tashlhiyt Corpus for Orthography Normalization and Low-Resource Language Processing.](https://md.research.elamur.ai/free/en/papers/monir-2026-datashi-parallel-english-tashlhiyt-corpus/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/26/)

## References

20 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Hemn Barzan Abdalla, Yulia Kumar, Jose Marchena, Stephany Guzman, Ardalan Awlla, Mehdi Gheisari, and Maryam Cheraghy. 2025. The future of artificial intelligence in the face of data scarcity. Computers, Materials and Continua, 84(1):1073–1099. [find](https://scholar.google.com/scholar?q=Hemn+Barzan+Abdalla%2C+Yulia+Kumar%2C+Jose+Marchena%2C+Stephany+Guzman%2C+Ardalan+Awlla%2C+Mehdi+Gheisari%2C+and+Maryam+Cheraghy%2E+2025%2E+The+future+of+artificial+intelligence+in+the+face+of+data+scarcity%2E+Computer)
2. Iman Barati, Mostafa Amiri, and Heshaam Faili. 2025. Searchinstruct: Enhancing domain adaptation via retrieval-based instruction dataset creation. arXiv preprint arXiv:2509.10708. — [arXiv:2509.10708](https://arxiv.org/abs/2509.10708) [find](https://scholar.google.com/scholar?q=Iman+Barati%2C+Mostafa+Amiri%2C+and+Heshaam+Faili%2E+2025%2E+Searchinstruct%3A+Enhancing+domain+adaptation+via+retrieval%2Dbased+instruction+dataset+creation%2E+arXiv+preprint+arXiv%3A2509%2E10708%2E+%E2%80%94+%5BarXiv%3A2509%2E10708%5D)
3. Daoyuan Chen, Yilun Huang, Zhijian Ma, Hesen Chen, Xuchen Pan, Ce Ge, Dawei Gao, Yuexiang Xie, Zhaoyang Liu, Jinyang Gao, and 1 others. 2024. Data-juicer: A one-stop data processing system for large language models. In Companion of the 2024 International Conference on Management of Data, pages 120–1… [find](https://scholar.google.com/scholar?q=Daoyuan+Chen%2C+Yilun+Huang%2C+Zhijian+Ma%2C+Hesen+Chen%2C+Xuchen+Pan%2C+Ce+Ge%2C+Dawei+Gao%2C+Yuexiang+Xie%2C+Zhaoyang+Liu%2C+Jinyang+Gao%2C+and+1+others%2E+2024%2E+Data%2Djuicer%3A+A+one%2Dstop+data+processing+system+for+large+l)
4. Tommaso Ferracci, Leonie Tabea Goldmann, Anton Hinel, and Francesco Sanna Passino. 2024. Targeted synthetic data generation for tabular data via hardness characterization. arXiv preprint arXiv:2410.00759. — [arXiv:2410.00759](https://arxiv.org/abs/2410.00759) [find](https://scholar.google.com/scholar?q=Tommaso+Ferracci%2C+Leonie+Tabea+Goldmann%2C+Anton+Hinel%2C+and+Francesco+Sanna+Passino%2E+2024%2E+Targeted+synthetic+data+generation+for+tabular+data+via+hardness+characterization%2E+arXiv+preprint+arXiv%3A2410%2E00)
5. Yoonsik Kim, Moonbin Yim, and Ka Yeon Song. 2024. Tablevqa-bench: A visual question answering benchmark on multiple table domains. arXiv preprint arXiv:2404.19205. — [arXiv:2404.19205](https://arxiv.org/abs/2404.19205) [find](https://scholar.google.com/scholar?q=Yoonsik+Kim%2C+Moonbin+Yim%2C+and+Ka+Yeon+Song%2E+2024%2E+Tablevqa%2Dbench%3A+A+visual+question+answering+benchmark+on+multiple+table+domains%2E+arXiv+preprint+arXiv%3A2404%2E19205%2E+%E2%80%94+%5BarXiv%3A2404%2E19205%5D%28https%3A%2F%2Farxiv%2Eo)
6. Haitao Li, You Chen, Qingyao Ai, Yueyue Wu, Ruizhe Zhang, and Yiqun Liu. 2024. Lexeval: A comprehensive chinese legal benchmark for evaluating large language models. Advances in Neural Information Processing Systems, 37:25061–25094. [find](https://scholar.google.com/scholar?q=Haitao+Li%2C+You+Chen%2C+Qingyao+Ai%2C+Yueyue+Wu%2C+Ruizhe+Zhang%2C+and+Yiqun+Liu%2E+2024%2E+Lexeval%3A+A+comprehensive+chinese+legal+benchmark+for+evaluating+large+language+models%2E+Advances+in+Neural+Information+Pro)
7. Hao Liang, Xiaochen Ma, Zhou Liu, Zhen Hao Wong, Zhengyang Zhao, Zimo Meng, Runming He, Chengyu Shen, Qifeng Cai, Zhaoyang Han, and 1 others. 2025. Dataflow: An llm-driven framework for unified data preparation and workflow automation in the era of data-centric ai. arXiv preprint arXiv:2512.16676. — [arXiv:2512.16676](https://arxiv.org/abs/2512.16676) [find](https://scholar.google.com/scholar?q=Hao+Liang%2C+Xiaochen+Ma%2C+Zhou+Liu%2C+Zhen+Hao+Wong%2C+Zhengyang+Zhao%2C+Zimo+Meng%2C+Runming+He%2C+Chengyu+Shen%2C+Qifeng+Cai%2C+Zhaoyang+Han%2C+and+1+others%2E+2025%2E+Dataflow%3A+An+llm%2Ddriven+framework+for+unified+data+p)
8. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. [G-eval: Nlg evaluation using gpt-4 with better human alignment](https://md.research.elamur.ai/free/en/papers/liu-2023-g-eval-nlg-evaluation-gpt/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 2511–2522.
9. Junyu Luo, Zhizhuo Kou, Liming Yang, Xiao Luo, Jinsheng Huang, Zhiping Xiao, Jingshu Peng, Chengzhong Liu, Jiaming Ji, Xuanzhe Liu, and 1 others. 2025. [Finmme: Benchmark dataset for financial multi-modal reasoning evaluation](https://md.research.elamur.ai/free/en/papers/luo-2025-finmme-benchmark-dataset-financial-multi/). In Proceedings of the 63rd Annual Meeting of the Association for Computati…
10. OpenAI. 2023. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). arXiv preprint arXiv:2303.08774.
11. Karim Ouda. 2025. ABBL: An Advanced Benchmark and Leaderboard for Comprehensive Evaluation of Arabic Language Models. Accessed: 2025-07-19. [find](https://scholar.google.com/scholar?q=Karim+Ouda%2E+2025%2E+ABBL%3A+An+Advanced+Benchmark+and+Leaderboard+for+Comprehensive+Evaluation+of+Arabic+Language+Models%2E+Accessed%3A+2025%2D07%2D19%2E)
12. Stephen Robertson, Hugo Zaragoza, and 1 others. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and trends® in information retrieval, 3(4):333–389. [find](https://scholar.google.com/scholar?q=Stephen+Robertson%2C+Hugo+Zaragoza%2C+and+1+others%2E+2009%2E+The+probabilistic+relevance+framework%3A+Bm25+and+beyond%2E+Foundations+and+trends%C2%AE+in+information+retrieval%2C+3%284%29%3A333%E2%80%93389%2E)
13. Yiduo Guo, Zhen Guo, Chuanwei Huang, Zi-Ang Wang, Zekai Zhang, Haofei Yu, Huishuai Zhang, and Yikang Shen. 2025. Synthetic data rl: Task definition is all you need. arXiv preprint arXiv:2505.17063. — [arXiv:2505.17063](https://arxiv.org/abs/2505.17063) [find](https://scholar.google.com/scholar?q=Yiduo+Guo%2C+Zhen+Guo%2C+Chuanwei+Huang%2C+Zi%2DAng+Wang%2C+Zekai+Zhang%2C+Haofei+Yu%2C+Huishuai+Zhang%2C+and+Yikang+Shen%2E+2025%2E+Synthetic+data+rl%3A+Task+definition+is+all+you+need%2E+arXiv+preprint+arXiv%3A2505%2E17063%2E+%E2%80%94+)
14. Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2021. [What disease does this patient have? a large-scale open domain question answering dataset from medical exams](https://md.research.elamur.ai/free/en/papers/jin-2020-disease-this-patient-have-large/). Applied Sciences, 11(14):6421.
15. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick SH Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. [Dense passage retrieval for open-domain question answering](https://md.research.elamur.ai/free/en/papers/karpukhin-2020-dense-passage-retrieval-open-domain/). In EMNLP (1), pages 6769–6781.
16. Omar Khattab and Matei Zaharia. 2020. [Colbert: Efficient and effective passage search via contextualized late interaction over bert](https://md.research.elamur.ai/free/en/papers/khattab-2020-colbert-efficient-effective-passage/). In Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval, pages 39– 48.
17. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, and 1 others. 2024. [Deepseekmath: Pushing the limits of mathematical reasoning in open language models](https://md.research.elamur.ai/free/en/papers/shao-2024-deepseekmath-pushing-limits-mathematical-reasoning/). arXiv preprint arXiv:2402.03300.
18. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2025. [Hybridflow: A flexible and efficient rlhf framework](https://md.research.elamur.ai/free/en/papers/sheng-2024-hybridflow-flexible-efficient-rlhf-framework/). In Proceedings of the Twentieth European Conference on Computer Systems, pages 1279–1297.
19. Yuxuan Tong, Xiwen Zhang, Rui Wang, Ruidong Wu, and Junxian He. 2024. Dart-math: Difficulty-aware rejection tuning for mathematical problem-solving. Advances in Neural Information Processing Systems, 37:7821–7846. [find](https://scholar.google.com/scholar?q=Yuxuan+Tong%2C+Xiwen+Zhang%2C+Rui+Wang%2C+Ruidong+Wu%2C+and+Junxian+He%2E+2024%2E+Dart%2Dmath%3A+Difficulty%2Daware+rejection+tuning+for+mathematical+problem%2Dsolving%2E+Advances+in+Neural+Information+Processing+Systems%2C+)
20. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. [Llama: Open and efficient foundation language models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-open-efficient-foundation-language/). arXiv preprint arXiv:2302.13971.

*Updated 1 November 2026*