---
title: "ProUIE — paper by Liu et al. (2026) — summary and ideas"
description: "Many LLM-based universal information extraction (UIE) systems add extra signals on top of the standard annotations, such as… 21 references."
authors: ["Wenda Liu", "Song Zhigang", "Shuai Nie", "Guangyao Liu", "Lisung Chen", "Binyu Yang", "Yaran Chen", "Peng Zhou", "Hongzhen Wang", "Yuchen Liu", "Wenyue Hu", "Jiaming Xu", "Runyu Shi", "Ying Huang"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.1093/"
doi: "10.18653/v1/2026.findings-acl.1093"
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/liu-2026-prouie-macro-micro-progressive-learning/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/liu-2026-prouie-macro-micro-progressive-learning/)

[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/46/) / ProUIE: A Macro-to-Micro Progressive Learning Method for LLM-based Universal Information Extraction

# ProUIE: A Macro-to-Micro Progressive Learning Method for LLM-based Universal Information Extraction

**Authors:** Wenda Liu, Song Zhigang, Shuai Nie, Guangyao Liu, Lisung Chen, Binyu Yang, [Yaran Chen](https://md.research.elamur.ai/free/en/authors/yaran-chen/), [Peng Zhou](https://md.research.elamur.ai/free/en/authors/peng-zhou/), Hongzhen Wang, [Yuchen Liu](https://md.research.elamur.ai/free/en/authors/yuchen-liu/), Wenyue Hu, [Jiaming Xu](https://md.research.elamur.ai/free/en/authors/jiaming-xu/), Runyu Shi and Ying Huang · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/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.findings-acl.1093/) · [PDF](https://aclanthology.org/2026.findings-acl.1093.pdf) · [DOI 10.18653/v1/2026.findings-acl.1093](https://doi.org/10.18653/v1/2026.findings-acl.1093) · [OpenAlex](https://openalex.org/W7166869299)

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

## Summary

Many LLM-based universal information extraction (UIE) systems add extra signals on top of the standard annotations, such as schema hints, external knowledge, guidelines, or verification pipelines. These additions make training heavier, and the payoff is often small. ProUIE, from a Xiaomi-led team, instead changes only how the model learns from the original annotated data. Training moves through three progressively finer stages:

1. **Complete Modeling (CM):** supervised fine-tuning on all data, ordered by difficulty (NER, then RE, then EE).
2. **Streamlined Alignment (SA):** further tuning on a sampled, mostly RE/EE subset whose targets are compacted by dropping empty or redundant fields.
3. **Deep Exploration (DE):** reinforcement learning with GRPO, driven by a new Stepwise Fine-grained Reward (SFR). SFR scores outputs at several structural levels, from types down to individual slots.

Across 36 public benchmarks, a 4B-parameter ProUIE achieves higher average NER and RE scores than competitive models that were tuned on instructions. It also does well on a confidential production task, mobile query understanding. The work suggests that a well-designed training curriculum alone can replace auxiliary supervision. Event extraction gains, however, remain uneven.

## Key points

- **CM curriculum:** the first phase trains on NER only. The second phase mixes NER and RE at 2:8, and the third mixes NER, RE, and EE at 3:3:4.
- **SFR reward:** it combines weighted coverage on coarse units, a powered micro-F1 on fine units, and a penalty for mismatches. It is applied in successive GRPO phases for NER, then RE, then EE. In ablations, the combined coarse-to-fine reward beat coarse-only and fine-only variants on all three tasks.
- **Shorter outputs from SA:** on SciERC and SemEval, mean RE output length fell from 70/83/114 to 13/18/74 tokens (P50/P70/P99). The largest P99 output dropped from 161 to 126 tokens. Shortening was also measured for EE on ACE2005 and CASIE.
- **Production evaluation:** ProUIE (4B) achieved higher slot-level ExactAcc than Qwen2.5-3B/7B and Qwen3-4B/8B on six slots: body, action, environment, location, number, and source.
- **Limitations:** improvements on EE are inconsistent at the 4B scale. The GRPO stage also adds compute cost compared with plain supervised fine-tuning.

*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

- [Xiaomi](https://md.research.elamur.ai/free/en/companies/xiaomi/) — Company, Beijing, China · 14 works · 434 citations
- [Xi’an Jiaotong-Liverpool University](https://md.research.elamur.ai/free/en/universities/xi-an-jiaotong-liverpool-university/) — University, Suzhou, China · 11 works · 319 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 3,486 works

🧭 Topics (1): [🔎 Information extraction](https://md.research.elamur.ai/free/en/topics/information-extraction/) · 175 works

## Cite

Cite as: Wenda Liu, Song Zhigang, Shuai Nie, Guangyao Liu, Lisung Chen, Binyu Yang, Yaran Chen, Peng Zhou, Hongzhen Wang, Yuchen Liu, Wenyue Hu, Jiaming Xu, Runyu Shi and Ying Huang. 2026. ProUIE: A Macro-to-Micro Progressive Learning Method for LLM-based Universal Information Extraction. Findings of the Association for Computational Linguistics: ACL 2026.

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

```bibtex
@inproceedings{liu2026prouie,
  title = {{ProUIE: A Macro-to-Micro Progressive Learning Method for LLM-based Universal Information Extraction}},
  author = {Wenda Liu, Song Zhigang, Shuai Nie, Guangyao Liu, Lisung Chen, Binyu Yang, Yaran Chen, Peng Zhou, Hongzhen Wang, Yuchen Liu, Wenyue Hu, Jiaming Xu, Runyu Shi and Ying Huang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1093},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1093}
}
```

- [← previous Protecting Language Models Against Unauthorized Distillation through Trace Rewriting](https://md.research.elamur.ai/free/en/papers/ma-2026-protecting-language-models-against-unauthorized/)
- [next → ProverbIT - Easy to complete, hard to choose: A CALAMITA Challenge](https://md.research.elamur.ai/free/en/papers/mensa-2026-proverbit-easy-complete-hard-choose/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/46/)

## References

21 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Fan Bai, Junmo Kang, Gabriel Stanovsky, Dayne Freitag, Mark Dredze, and Alan Ritter. 2024. Schema-driven information extraction from heterogeneous tables. Preprint, arXiv:2305.14336. the Association for Computational Linguistics: ACL 2025, pages 2403–2414, Vienna, Austria. Association for Computatio… [find](https://scholar.google.com/scholar?q=Fan+Bai%2C+Junmo+Kang%2C+Gabriel+Stanovsky%2C+Dayne+Freitag%2C+Mark+Dredze%2C+and+Alan+Ritter%2E+2024%2E+Schema%2Ddriven+information+extraction+from+heterogeneous+tables%2E+Preprint%2C+arXiv%3A2305%2E14336%2E+the+Association+f)
2. Zixuan Li, Yutao Zeng, Yuxin Zuo, Weicheng Ren, Wenxuan Liu, Miao Su, Yucan Guo, Yantao Liu, Xiang Li, Zhilei Hu, Long Bai, Wei Li, Yidan Liu, Pan Yang, Xiaolong Jin, Jiafeng Guo, and Xueqi Cheng. 2024. [Knowcoder: Coding structured knowledge into llms for universal information extraction](https://md.research.elamur.ai/free/en/papers/li-2024-knowcoder-coding-structured-knowledge-into/). Preprint,…
3. Xincheng Liao, Junwen Duan, Yixi Huang, and Jianxin Wang. 2025. Ruie: Retrieval-based unified information extraction using large language model. Preprint, arXiv:2409.11673. [find](https://scholar.google.com/scholar?q=Xincheng+Liao%2C+Junwen+Duan%2C+Yixi+Huang%2C+and+Jianxin+Wang%2E+2025%2E+Ruie%3A+Retrieval%2Dbased+unified+information+extraction+using+large+language+model%2E+Preprint%2C+arXiv%3A2409%2E11673%2E)
4. Yankai Lin, Shiqi Shen, Zhiyuan Liu, Huanbo Luan, and Maosong Sun. 2016. Neural relation extraction with selective attention over instances. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2124–2133, Berlin, Germany. Associati… [find](https://scholar.google.com/scholar?q=Yankai+Lin%2C+Shiqi+Shen%2C+Zhiyuan+Liu%2C+Huanbo+Luan%2C+and+Maosong+Sun%2E+2016%2E+Neural+relation+extraction+with+selective+attention+over+instances%2E+In+Proceedings+of+the+54th+Annual+Meeting+of+the+Associatio)
5. Jian Liu, Yubo Chen, Kang Liu, Wei Bi, and Xiaojiang Liu. 2020. [Event extraction as machine reading comprehension](https://md.research.elamur.ai/free/en/papers/liu-2020-event-extraction-as-machine-reading/). In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1641–1651, Online. Association for Computational Linguistics.
6. Yaojie Lu, Qing Liu, Dai Dai, Xinyan Xiao, Hongyu Lin, Xianpei Han, Le Sun, and Hua Wu. 2022. Unified structure generation for universal information extraction. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5755–5772, Dublin… [find](https://scholar.google.com/scholar?q=Yaojie+Lu%2C+Qing+Liu%2C+Dai+Dai%2C+Xinyan+Xiao%2C+Hongyu+Lin%2C+Xianpei+Han%2C+Le+Sun%2C+and+Hua+Wu%2E+2022%2E+Unified+structure+generation+for+universal+information+extraction%2E+In+Proceedings+of+the+60th+Annual+Meeti)
7. Xuezhe Ma and Eduard Hovy. 2016. [End-to-end sequence labeling via bi-directional lstm-cnns-crf](https://md.research.elamur.ai/free/en/papers/ma-2016-end-end-sequence-labeling-bi/). Preprint, arXiv:1603.01354.
8. Hao Fei, Shengqiong Wu, Jingye Li, Bobo Li, Fei Li, Libo Qin, Meishan Zhang, Min Zhang, and Tat-Seng Chua. 2023. Lasuie: Unifying information extraction with latent adaptive structure-aware generative language model. Preprint, arXiv:2304.06248. — [arXiv:2304.06248](https://arxiv.org/abs/2304.06248) [find](https://scholar.google.com/scholar?q=Hao+Fei%2C+Shengqiong+Wu%2C+Jingye+Li%2C+Bobo+Li%2C+Fei+Li%2C+Libo+Qin%2C+Meishan+Zhang%2C+Min+Zhang%2C+and+Tat%2DSeng+Chua%2E+2023%2E+Lasuie%3A+Unifying+information+extraction+with+latent+adaptive+structure%2Daware+generative)
9. Chaoxu Pang, Yixuan Cao, Qiang Ding, and Ping Luo. 2023. Guideline learning for in-context information extraction. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 15372–15389, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Chaoxu+Pang%2C+Yixuan+Cao%2C+Qiang+Ding%2C+and+Ping+Luo%2E+2023%2E+Guideline+learning+for+in%2Dcontext+information+extraction%2E+In+Proceedings+of+the+2023+Conference+on+Empirical+Methods+in+Natural+Language+Proces)
10. Yucan Guo, Zixuan Li, Xiaolong Jin, Yantao Liu, Yutao Zeng, Wenxuan Liu, Xiang Li, Pan Yang, Long Bai, Jiafeng Guo, and Xueqi Cheng. 2023. Retrieval-augmented code generation for universal information extraction. Preprint, arXiv:2311.02962. — [arXiv:2311.02962](https://arxiv.org/abs/2311.02962) [find](https://scholar.google.com/scholar?q=Yucan+Guo%2C+Zixuan+Li%2C+Xiaolong+Jin%2C+Yantao+Liu%2C+Yutao+Zeng%2C+Wenxuan+Liu%2C+Xiang+Li%2C+Pan+Yang%2C+Long+Bai%2C+Jiafeng+Guo%2C+and+Xueqi+Cheng%2E+2023%2E+Retrieval%2Daugmented+code+generation+for+universal+information)
11. Oscar Sainz, Iker García-Ferrero, Rodrigo Agerri, Oier Lopez de Lacalle, German Rigau, and Eneko Agirre. 2024. [Gollie: Annotation guidelines improve zero-shot information-extraction](https://md.research.elamur.ai/free/en/papers/sainz-2023-gollie-annotation-guidelines-improve-zero/). Preprint, arXiv:2310.03668.
12. Guillaume Lample, Miguel Ballesteros, Sandeep Subramanian, Kazuya Kawakami, and Chris Dyer. 2016. [Neural architectures for named entity recognition](https://md.research.elamur.ai/free/en/papers/lample-2016-neural-architectures-named-entity-recognition/). Preprint, arXiv:1603.01360.
13. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 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/). Preprint, arXiv:2402.03300.
14. Bo Li, Gexiang Fang, Wei Ye, Zhenghua Xu, Jinglei Zhang, Hao Cheng, and Shikun Zhang. 2025. MPL: Multiple programming languages with large language models for information extraction. In Findings of Zhen Wan, Fei Cheng, Zhuoyuan Mao, Qianying Liu, Haiyue Song, Jiwei Li, and Sadao Kurohashi. 2023. Gpt… [find](https://scholar.google.com/scholar?q=Bo+Li%2C+Gexiang+Fang%2C+Wei+Ye%2C+Zhenghua+Xu%2C+Jinglei+Zhang%2C+Hao+Cheng%2C+and+Shikun+Zhang%2E+2025%2E+MPL%3A+Multiple+programming+languages+with+large+language+models+for+information+extraction%2E+In+Findings+of+Zh)
15. Xiao Wang, Weikang Zhou, Can Zu, Han Xia, Tianze Chen, Yuansen Zhang, Rui Zheng, Junjie Ye, Qi Zhang, Tao Gui, Jihua Kang, Jingsheng Yang, Siyuan Li, and Chunsai Du. 2023. [Instructuie: Multi-task instruction tuning for unified information extraction](https://md.research.elamur.ai/free/en/papers/wang-2023-instructuie-multi-task-instruction-tuning/). Preprint, arXiv:2304.08085.
16. Xiang Wei, Xingyu Cui, Ning Cheng, Xiaobin Wang, Xin Zhang, Shen Huang, Pengjun Xie, Jinan Xu, Yufeng Chen, Meishan Zhang, Yong Jiang, and Wenjuan Han. 2024. [Chatie: Zero-shot information extraction via chatting with chatgpt](https://md.research.elamur.ai/free/en/papers/wei-2023-chatie-zero-shot-information-extraction/). Preprint, arXiv:2302.10205.
17. Zhihao Wen, Sheng Liang, Yaxiong Wu, Yongyue Zhang, and Yong Liu. 2025. Effective and efficient schema-aware information extraction using on-device large language models. Preprint, arXiv:2505.14992. — [arXiv:2505.14992](https://arxiv.org/abs/2505.14992) [find](https://scholar.google.com/scholar?q=Zhihao+Wen%2C+Sheng+Liang%2C+Yaxiong+Wu%2C+Yongyue+Zhang%2C+and+Yong+Liu%2E+2025%2E+Effective+and+efficient+schema%2Daware+information+extraction+using+on%2Ddevice+large+language+models%2E+Preprint%2C+arXiv%3A2505%2E14992%2E+%E2%80%94)
18. Xinglin Xiao, Yijie Wang, Nan Xu, Yuqi Wang, Hanxuan Yang, Minzheng Wang, Yin Luo, Lei Wang, Wenji Mao, and Daniel Zeng. 2024. Yayi-uie: A chat-enhanced instruction tuning framework for universal information extraction. Preprint, arXiv:2312.15548. — [arXiv:2312.15548](https://arxiv.org/abs/2312.15548) [find](https://scholar.google.com/scholar?q=Xinglin+Xiao%2C+Yijie+Wang%2C+Nan+Xu%2C+Yuqi+Wang%2C+Hanxuan+Yang%2C+Minzheng+Wang%2C+Yin+Luo%2C+Lei+Wang%2C+Wenji+Mao%2C+and+Daniel+Zeng%2E+2024%2E+Yayi%2Duie%3A+A+chat%2Denhanced+instruction+tuning+framework+for+universal+info)
19. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). Preprint, arXiv:2505.09388.
20. Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. 2024. [Llamafactory: Unified efficient fine-tuning of 100+ language models](https://md.research.elamur.ai/free/en/papers/zheng-2024-llamafactory-unified-efficient-fine-tuning/). Preprint, arXiv:2403.13372.
21. Wenxuan Zhou, Sheng Zhang, Yu Gu, Muhao Chen, and Hoifung Poon. 2024. [Universalner: Targeted distillation from large language models for open named entity recognition](https://md.research.elamur.ai/free/en/papers/zhou-2023-universalner-targeted-distillation-large-language/). Preprint, arXiv:2308.03279.

*Updated 2 November 2026*