Elamur Research⌕
☰
home / papers / 2026 / YNU-HPCC at SemEval-2026 Task 12: Retrieval-Guided Reasoning with Teacher Distillation for Abductive Event Reasoning

YNU-HPCC at SemEval-2026 Task 12: Retrieval-Guided Reasoning with Teacher Distillation for Abductive Event Reasoning

Paper · 2026 · Yuwei Sun, Jin Wang and Xuejie Zhang

Summary

This paper presents the YNU-HPCC system submitted to SemEval-2026's Task 12 on Abductive Event Reasoning (AER). The task asks a system to select every option that directly causes a target event, using multi-document retrieved evidence that includes distractors. Scoring works on sets: an exact match earns full credit, a non-empty proper subset earns half, and anything else earns zero. The authors treat the task as multi-label classification over four options and build a two-stage pipeline. In Stage 1, a DeBERTa-v3-base student processes long documents as overlapping windows. BM25 ranks these windows against a query that combines the event and options, and Top-K mean pooling aggregates the window scores. The student is trained with Focal Loss, paraphrase augmentation and hard negatives. In Stage 2, the student is fine-tuned on soft labels from a Qwen-14B teacher, keeping only confident pseudo-labels. Dev results are near ceiling, but test topics are unseen and cause strong domain shift. The full pipeline raises the test score from 0.46 to 0.73. The results suggest that evidence coverage and pseudo-label filtering matter more than encoder strength alone.

Key points

  • Stage 1 reaches an official dev score of 0.9712 (micro-F1 0.9746, macro-F1 0.9745), compared with 0.5437 for a BERT single-choice baseline. The final test score of 0.73 ranks 84th of 221 submissions.
  • Test topics and documents have zero overlap with train/dev, and the longest train/dev document is about 280,000 characters.
  • Settings: window 384, stride 224, max_len 512, up to 32 windows per question, Top-K with K=3, and predicted sets capped at 3 options.
  • Ablations:
    • Swapping BERT for DeBERTa without evidence modeling lowered the score (0.5437 → 0.5013).
    • Sliding windows, Top-K pooling and Focal Loss gave the main gains.
    • R-Drop and positive-class reweighting hurt the DeBERTa route.
  • Test trajectory: Stage 1 scored 0.46. Qwen2.5-3B-Instruct paraphrasing raised it to 0.57, hard negatives to 0.63, and unfiltered distillation to 0.65. High-confidence filtering gave the final 0.73, which coincided with smaller predicted sets.

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

🎓Yunnan UniversityKunming · China17 works · 40 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 worksInformation Systems · 1,437 works

🧭 Topics 3Research subjects and languages the work is about.Reasoning · 1,426 worksInformation retrieval & search · 473 worksEfficient models & compression · 242 works

Cite

Cite as Yuwei Sun, Jin Wang and Xuejie Zhang. 2026. YNU-HPCC at SemEval-2026 Task 12: Retrieval-Guided Reasoning with Teacher Distillation for Abductive Event Reasoning. Proceedings of the 20th International Workshop on Semantic Evaluation (2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.semeval-1.30

@inproceedings{sun2026ynu,
  title = {{YNU-HPCC at SemEval-2026 Task 12: Retrieval-Guided Reasoning with Teacher Distillation for Abductive Event Reasoning}},
  author = {Yuwei Sun, Jin Wang and Xuejie Zhang},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.30},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.30}
}

References

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

  1. Chandra Bhagavatula, Ronan Le Bras, Chaitanya Malaviya, Keisuke Sakaguchi, Ari Holtzman, Hannah Rashkin, Doug Downey, Scott Wen tau Yih, and Yejin Choi. 2020. Abductive commonsense reasoning. International Conference on Learning Representations (ICLR).
  2. Antoine Bosselut, Hannah Rashkin, Maarten Sap, Chaitanya Malaviya, Asli Celikyilmaz, and Yejin Choi. 2019. COMET: Commonsense transformers for automatic knowledge graph construction. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4762–4779. Associat…
  3. Chunkit Chan, Xin Liu, Tsz Ho Chan, Jiayang Cheng, Yangqiu Song, Ginny Wong, and Simon See. 2023. Self-consistent narrative prompts on abductive natural language inference. find
  4. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North, pages 4171–4186. Association for Computational Linguistics.
  5. Bosheng Ding, Chengwei Qin, Ruochen Zhao, Tianze Luo, Xinze Li, Guizhen Chen, Wenhan Xia, Junjie Hu, Anh Tuan Luu, and Shafiq Joty. 2024. Data augmentation using LLMs: Data perspectives, learning paradigms and challenges. In Findings of the Association for Computational Linguistics ACL 2024, pages 1… find
  6. Shailja Gupta, Rajesh Ranjan, and Surya Narayan Singh. 2024. A comprehensive survey of retrieval-augmented generation (RAG): Evolution, current landscape and future directions. arXiv. find
  7. Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. Debertav3: Improving deberta using electra-style pre-training with gradient-disentangled embedding sharing.
  8. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the knowledge in a neural network. arXiv.
  9. Cheng-Yu Hsieh, Chun-Liang Li, Chih kuan Yeh, Hootan Nakhost, Yasuhisa Fujii, Alex Ratner, Ranjay Krishna, Chen-Yu Lee, and Tomas Pfister. 2023. Distilling step-by-step! outperforming larger language models with less training data and smaller model sizes. In Findings of the Association for Computati…
  10. Pengcheng Jiang, Siru Ouyang, Yizhu Jiao, Ming Zhong, Runchu Tian, and Jiawei Han. 2025. Retrieval and structuring augmented generation with large language models. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, volume 2, pages 6032– 6042. Associatio… find
  11. Juanhui Li, Sreyashi Nag, Hui Liu, Xianfeng Tang, Sheikh Muhammad Sarwar, Limeng Cui, Hansu Gu, Suhang Wang, Qi He, and Jiliang Tang. 2025. Learning with less: Knowledge distillation from large language models via unlabeled data. In Findings of the Association for Computational Linguistics: NAACL 20… find
  12. Xiaobo Liang, Lijun Wu, Juntao Li, Yue Wang, Qi Meng, Tao Qin, Wei Chen, Min Zhang, and Tie-Yan Liu. 2021. R-drop: Regularized dropout for neural networks. arXiv. find
  13. Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollar. 2017. Focal loss for dense object detection. In 2017 IEEE International Conference on Computer Vision (ICCV), pages 2999–3007. IEEE.
  14. Hansa Meghwani, Amit Agarwal, Priyaranjan Pattnayak, Hitesh Laxmichand Patel, and Srikant Panda. 2025. Hard negative mining for domain-specific retrieval in enterprise systems. find
  15. Paulius Micikevicius, Sharan Narang, Jonah Alben, Gregory Diamos, Erich Elsen, David Garcia, Boris Ginsburg, Michael Houston, Oleksii Kuchaiev, Ganesh Venkatesh, and Hao Wu. 2018. Mixed precision training. find
  16. Lianhui Qin, Vered Shwartz, Peter West, Chandra Bhagavatula, Jena D. Hwang, Ronan Le Bras, Antoine Bosselut, and Yejin Choi. 2020. Back to the future: Unsupervised backprop-based decoding for counterfactual and abductive commonsense reasoning. In Proceedings of the 2020 Conference on Empirical Metho…
  17. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. arXiv. find
  18. Maarten Sap, Ronan LeBras, Emily Allaway, Chandra Bhagavatula, Nicholas Lourie, Hannah Rashkin, Brendan Roof, Noah A. Smith, and Yejin Choi. 2019. ATOMIC: An atlas of machine commonsense for if-then reasoning. Proceedings of AAAI 2019. find
  19. Xiaohua Wang, Zhenghua Wang, Xuan Gao, Feiran Zhang, Yixin Wu, Zhibo Xu, Tianyuan Shi, Zhengyuan Wang, Shizheng Li, Qi Qian, Ruicheng Yin, Changze Lv, Xiaoqing Zheng, and Xuanjing Huang. 2024. Searching for best practices in retrieval-augmented generation. In Proceedings of the 2024 Conference on Em…
  20. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. Self-instruct: Aligning language models with self-generated instructions. arXiv. find
  21. Xiaohan Xu, Ming Li, Chongyang Tao, Tao Shen, Reynold Cheng, Jinyang Li, Can Xu, Dacheng Tao, and Tianyi Zhou. 2024. A survey on knowledge distillation of large language models. arXiv. find
  22. Jiachen Zhao, Wenlong Zhao, Andrew Drozdov, Benjamin Rozonoyer, Md Arafat Sultan, Jay-Yoon Lee, Mohit Iyyer, and Andrew McCallum. 2024. Multistage collaborative knowledge distillation from a large language model for semi-supervised sequence generation. In Proceedings of the 62nd Annual Meeting of th… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/sun-2026-ynu-hpcc-semeval-2026-task/