Elamur Research⌕
☰
home / papers / 2026 / DEBAR: Mitigating Contextual Bias in Cross-Document Relation Extraction via Dual-Stream Decoupling

DEBAR: Mitigating Contextual Bias in Cross-Document Relation Extraction via Dual-Stream Decoupling

Paper · 2026 · Zhixuan Yang, Fu Zhang, Huangming Xu et al.

Summary

Cross-document relation extraction (CodRE) predicts the relation between a head and a tail entity that appear in different documents, linked through shared bridge entities. The authors argue that current systems merge target and bridge entities into one representation. This lets evidence tied to only one side dominate, a "relation transfer bias" that breaks reasoning chains. These systems also use a single global threshold that does not ensure positive relations outscore negative ones.

DEBAR addresses both problems in three parts:

  • Bridge-aware input construction scores bridge entities by how often they co-occur with both the head and the tail at sentence, paragraph and document level. It then builds separate head and tail context streams.
  • A dual-stream encoder feeds an entity graph processed by a Graph Recurrent Network.
  • A curriculum-aware ranking loss (CAO) gradually narrows the threshold margin and penalises negatives that rank above positives.

On CodRED, DEBAR reaches state-of-the-art results under closed and open settings. CAO also improves other backbones, which suggests it can serve as a general replacement for CodRE training losses.

Key points

  • Bridge scoring: connection strength is a weighted sum of sentence-, paragraph- and document-level co-occurrence counts. A hierarchical exclusion rule prevents the same co-occurrence from being counted twice.
  • CAO generalisation: adding CAO raised F1 from 51.26 to 52.37 for the End-to-end baseline and from 60.85 to 61.73 for ECRIM.
  • Comparison with LLMs: under the NEPD protocol (CodRED dev set, closed setting, Micro-F1), DEBAR scores 84.63. This beats InstructUIE-FT (80.72) by 3.91 points and far exceeds GPT-3.5-turbo (28.05).
  • Case study: on the triple (Sexxx Dreams, Synth-pop, genre), ECRIM picks a noisy tail-side sentence. DEBAR instead retrieves the sentence that links the bridge entity Lady Gaga to synth-pop.
  • Limitations and code: the separate encoders add slight training latency, and the method depends on explicit bridge entities. Code is available at github.com/newyuyou/DEBAR.

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

 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 Zhixuan Yang, Fu Zhang, Huangming Xu and Jingwei Cheng. 2026. DEBAR: Mitigating Contextual Bias in Cross-Document Relation Extraction via Dual-Stream Decoupling. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1091

@inproceedings{yang2026debar,
  title = {{DEBAR: Mitigating Contextual Bias in Cross-Document Relation Extraction via Dual-Stream Decoupling}},
  author = {Zhixuan Yang, Fu Zhang, Huangming Xu and Jingwei Cheng},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1091},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1091}
}

References

17 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Xu Han, Tianyu Gao, Yankai Lin, Hao Peng, Yaoliang Yang, Chaojun Xiao, Zhiyuan Liu, Peng Li, Jie Zhou, and Maosong Sun. 2020. More data, more relations, more context and more openness: A review and outlook for relation extraction. In Proceedings of the 1st Conference of the Asia-Pacific Chapter of t… find
  2. Monika Jain, Raghava Mutharaju, Kuldeep Singh, and Ramakanth Kavuluru. 2024. Knowledge-driven cross-document relation extraction. In Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024, pages 3787–3797. Association for Comput… find
  3. Bo Li, Wei Ye, Zhonghao Sheng, Rui Xie, Xiangyu Xi, and Shikun Zhang. 2020. Graph enhanced dual attention network for document-level relation extraction. In Proceedings of the 28th International Conference on Computational Linguistics, COLING 2020, Barcelona, Spain (Online), December 8-13, 2020, pag… find
  4. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net. find
  5. Chonggang Lu, Richong Zhang, Kai Sun, Jaein Kim, Cunwang Zhang, and Yongyi Mao. 2023a. Anaphor assisted document-level relation extraction. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 15453– 15464. Assoc… find
  6. Keming Lu, I-Hung Hsu, Wenxuan Zhou, Mingyu Derek Ma, and Muhao Chen. 2023b. Multi-hop evidence retrieval for cross-document relation extraction. In Findings of the Association for Computational Linguistics: ACL 2023, Toronto, Canada, July 9-14, 2023, pages 10336–10351. Association for Computational… find
  7. Youmi Ma, An Wang, and Naoaki Okazaki. 2023. DREEAM: guiding attention with evidence for improving document-level relation extraction. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023, Dubrovnik, Croatia, May 2-6, 2023, pages 1… find
  8. Byeonghu Na, Suhyeon Jo, Yeongmin Kim, and Il-Chul Moon. 2024. Reward-based input construction for cross-document relation extraction. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok, Thailand, August 11-16, 2024,… find
  9. Fengqi Wang, Fei Li, Hao Fei, Jingye Li, Shengqiong Wu, Fangfang Su, Wenxuan Shi, Donghong Ji, and Bo Cai. 2022. Entity-centered cross-document relation extraction. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, EMNLP 2022, Abu Dhabi, United Arab Emirates,… find
  10. Linlin Wang, Zhu Cao, Gerard de Melo, and Zhiyuan Liu. 2016. Relation classification via multi-level attention cnns. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, ACL 2016, August 7-12, 2016, Berlin, Germany, Volume 1: Long Papers. The Association for Co…
  11. Haoran Wu, Xiuyi Chen, Zefa Hu, Jing Shi, Shuang Xu, and Bo Xu. 2023. Local-to-global causal reasoning for cross-document relation extraction. IEEE CAA J. Autom. Sinica, 10(7):1608–1621. find
  12. Lilong Xue, Dan Zhang, Yuxiao Dong, and Jie Tang. 2024. Autore: Document-level relation extraction with large language models. CoRR, abs/2403.14888.
  13. Yuan Yao, Jiaju Du, Yankai Lin, Peng Li, Zhiyuan Liu, Jie Zhou, and Maosong Sun. 2021. Codred: A cross-document relation extraction dataset for acquiring knowledge in the wild. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021, Virtual Event / Punt… find
  14. Yuan Yao, Deming Ye, Peng Li, Xu Han, Yankai Lin, Zhenghao Liu, Zhiyuan Liu, Lixin Huang, Jie Zhou, and Maosong Sun. 2019. Docred: A large-scale document-level relation extraction dataset. In Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Ita… find
  15. Hao Yue, Shaopeng Lai, Chengyi Yang, Liang Zhang, Junfeng Yao, and Jinsong Su. 2024. Towards better graph-based cross-document relation extraction via non-bridge entity enhancement and prediction debiasing. In Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and… find
  16. Klim Zaporojets, Johannes Deleu, Chris Develder, and Thomas Demeester. 2021. DWIE: an entity-centric dataset for multi-task document-level information extraction. Inf. Process. Manag., 58(4):102563. find
  17. Shuang Zeng, Runxin Xu, Baobao Chang, and Lei Li. 2020. Double graph based reasoning for document-level relation extraction. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online, November 16-20, 2020, pages 1630–1640. Association for Computati…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2026-debar-mitigating-contextual-bias-cross/