Elamur Research⌕
☰
home / papers / 2026 / A Self-Reflective LLM-based Architecture for Semi-Open Event Extraction

A Self-Reflective LLM-based Architecture for Semi-Open Event Extraction

Paper · 2026 · Hristo Tanev, Michel de Bollivier and Bertrand De Longueville

Summary

Event extraction usually involves a trade-off. Closed schemas such as ACE or ERE keep outputs consistent but miss information outside the template. Open event extraction is flexible but can yield inconsistent records. The authors, from the European Commission's Joint Research Centre, propose Semi-Open Event Extraction (SOEE). A user-defined core of fields (event type, date, location and others) stays fixed, and further attributes are added at runtime depending on the article.

Their prototype, MAREA, is a three-layer multi-agent system built on a generative LLM:

  • Expert layer: builds a few-shot initial template and answers questions.
  • Reflective layer: asks itself about missing event details and proposes new fields.
  • Coordination layer: a manager agent controls the workflow.

On health-related news articles, MAREA extracted core fields well, and most reflection-added fields were judged useful. Errors came mostly from misread prompts or misread articles rather than hallucination. The approach matters because it offers comparable records for event databases while still capturing context-specific detail, and its reliable fields could seed silver-standard corpora.

Key points

  • Reflection mechanism: an internal question-generation and answering loop with three question strategies, one of them keyword-based. A spatial reasoning agent uses spaCy to find place names, and the LLM then assigns each place a role (e.g., place of infection), creating new candidate fields.
  • Core-field F1: high for where-country (0.986), measures taken (0.941), actors (0.931) and event_type (0.901). Weaker results were for where-place (0.560) and when-end (0.462).
  • Quality of reflection-added fields: 36% high, 20% medium, 12% low, 5% irrelevant and 24% duplicates.
  • BERT-suggested questions: micro precision 0.7201, recall 0.4381 and F1 0.5448, which corresponds to a miss probability of about 0.4 in a small-scale test.
  • Limitations: a single annotator, small samples and one domain. Planned work includes prompt optimization (e.g., MIPROv2), post-processing filters, other LLM backends, and new domains such as security and disaster relief.

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

🏛️European CommissionBrussels · Belgium5 works · 974 citations🏛️Joint Research CentreIspra · Italy5 works · 974 citations🏛️Joint Research CentrePetten · Netherlands3 works · 815 citations

Also on the byline: Joint Research Centre.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.Information extraction · 175 works

Cite

Cite as Hristo Tanev, Michel de Bollivier and Bertrand De Longueville. 2026. A Self-Reflective LLM-based Architecture for Semi-Open Event Extraction. Proceedings of the 9th Workshop on Event Extraction and Understanding: Challenges and Applications (EEUCA 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eeuca-1.8

@inproceedings{tanev2026self,
  title = {{A Self-Reflective LLM-based Architecture for Semi-Open Event Extraction}},
  author = {Hristo Tanev, Michel {de Bollivier} and Bertrand {De Longueville}},
  booktitle = {Proceedings of the 9th Workshop on Event Extraction and Understanding: Challenges and Applications (EEUCA 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.eeuca-1.8},
  url = {https://doi.org/10.18653/v1/2026.eeuca-1.8}
}

References

26 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. 1662–1675, New Orleans, Louisiana. Association for Computational Linguistics. find
  2. Chinatsu Aone and Mila Ramos-Santacruz. 2000. Rees: a large-scale relation and event extraction system. In Sixth applied natural language processing conference, pages 76–83. find
  3. Haolin Deng, Yanan Zhang, Yangfan Zhang, Wangyang Ying, Changlong Yu, Jun Gao, Wei Wang, Xiaoling Bai, Nan Yang, Jin Ma, and 1 others. 2022. 2event: Benchmarking open event extraction with a large-scale chinese title dataset. In Proceedings of the 2022 Conference on Empirical Methods in Natural Lang… find
  4. Asli Celikyilmaz, Antoine Bosselut, Xiaodong He, and Yejin Choi. 2018. Deep communicating agents for abstractive summarization. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Long Papers), pages Ralph…
  5. Krista Opsahl-Ong, Michael J Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia, and Omar Khattab. 2024. Optimizing instructions and demonstrations for multi-stage language model programs. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, page…
  6. Quanjiang Guo, Sijie Wang, Jinchuan Zhang, Ben Zhang, Zhao Kang, Ling Tian, and Ke Yan. 2026. Extracting events like code: A multi-agent programming framework for zero-shot event extraction. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 30880–30887. find
  7. Jakub Piskorski, Nicolas Stefanovitch, Brian Doherty, Jens P Linge, Sopho Kharazi, Jas Mantero, Guillaume Jacquet, Alessio Spadaro, Giulia Teodori, and 1 others. 2023. Multi-label infectious disease news event corpus. In Text2Story@ ECIR, pages 171–183. find
  8. Zijin Hong and Jian Liu. 2024. Towards better question generation in QA-based event extraction. In Findings of the Association for Computational Linguistics: ACL 2024, pages 9025–9038, Bangkok, Thailand. Association for Computational Linguistics. find
  9. Zhiyi Song, Ann Bies, Stephanie Strassel, Tom Riese, Justin Mott, Joe Ellis, Jonathan Wright, Seth Kulick, Neville Ryant, and Xiaoyi Ma. 2015. From light to rich ere: Annotation of entities, relations, and events. In Proceedings of the 3rd workshop on EVENTS: Definition, detection, coreference, and…
  10. Hyuntak Kim and Byung-Hak Kim. 2025. Nexus-Sum: Hierarchical LLM agents for long-form narrative summarization. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10120– 10157, Vienna, Austria. Association for Computational Lingui… find
  11. Ralf Steinberger, Bruno Pouliquen, and Erik Van der Goot. 2013. An introduction to the europe media monitor family of applications. Information Access in a Multilingual World. find
  12. Bobo Li, Xudong Han, Jiang Liu, Yuzhe Ding, Liqiang Jing, Zhaoqi Zhang, Jinheng Li, Xinya Du, Fei Li, Meishan Zhang, and 1 others. 2025. Event extraction in large language model:a holistic survey of method, modality, and future. arXiv preprint arXiv:2512.19537. — [arXiv:2512.19537](https://arxiv.org/abs/2512.19537) find
  13. Hristo Tanev, Nicolas Stefanovitch, Tomáš Harmatha, and Diana F. Sousa. 2025a. Exploring the performance of large language models for event detection and extraction in the health domain. In Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing - Natural L… find
  14. Hongzhan Lin, Yang Deng, Yuxuan Gu, Wenxuan Zhang, Jing Ma, See Kiong Ng, and Tat-Seng Chua. 2025. Fact-audit: An adaptive multi-agent framework for dynamic fact-checking evaluation of large language models. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (…
  15. Hristo Tanev, Nicolas Stefanovitch, Tomáš Harmatha, and Diana F Sousa. 2025b. Exploring the performance of large language models for event detection and extraction in the health domain. In Proceedings of the International Conference on Recent Advances in Natural Language Processing. find
  16. Jens P Linge, Marco Verile, Hristo Tanev, Vanni Zavarella, Flavio Fuart, and Erik van der Goot. 2012. Media monitoring of public health threats with medisys. C. WILLIAM, CWR. WEB-STER, D. BALAHUR, et al, pages 17–31. find
  17. Di Lu, Shihao Ran, Joel Tetreault, and Alejandro Jaimes. 2023. Event extraction as question generation and answering. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 1666–1688, Toronto, Canada. Association for Computational L… find
  18. Meihan Tong, Bin Xu, Shuai Wang, Yixin Cao, Lei Hou, Juanzi Li, and Jun Xie. 2020. Improving event detection via open-domain event trigger knowledge. Association for Computational Linguistics. find
  19. Yuli Vasiliev. 2020. Natural Language Processing with Python and spaCy: A Practical Introduction. No Starch Press. find
  20. Meng Lu, Yuzhang Xie, Zhenyu Bi, Shuxiang Cao, and Xuan Wang. 2025. Crossagentie: Cross-type and cross-task multi-agent llm collaboration for zero-shot information extraction. In Findings of the Association for Computational Linguistics: ACL 2025. Association for Computational Linguistics. find
  21. Rui Wang, Jiaoli Liu, Yu Yan, Liwei Zang, Huimin Wang, and Jianyi Liu. 2025. Document-level event extraction framework based on prompt learning. In International Conference on Computer Application and Information Security (ICCAIS 2024), volume 13562, pages 996–1001. SPIE. find
  22. Zihao Meng, Tao Liu, Heng Zhang, Kai Feng, and Peng Zhao. 2024. Cean: Contrastive event aggregation network with llm-based augmentation for event extraction. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages… find
  23. Rui Wang, Deyu Zhou, and Yulan He. 2019. Open event extraction from online text using a generative adversarial network. arXiv preprint arXiv:1908.09246. find
  24. Sijia Wang and Lifu Huang. 2024. Debate as optimization: Adaptive conformal prediction and diverse retrieval for event extraction. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16422–16435, Miami, Florida, USA. Association for Computational Linguistics. find
  25. Guangjun Zhang, Hu Zhang, Yazhou Han, Yue Fan, Yuhang Shao, Hongye Tan, and Ru Li. 2026. Learning to generate and extract: A multi-agent collaboration framework for zero-shot document-level event arguments extraction. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages… find
  26. Chang Zong, Yuchen Yan, Weiming Lu, Jian Shao, Yongfeng Huang, Heng Chang, and Yueting Zhuang. 2024. Triad: A framework leveraging a multi-role LLM-based agent to solve knowledge base question answering. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tanev-2026-self-reflective-llm-based-architecture/