Elamur Research⌕
☰
home / papers / 2026 / DAL Team at PsyDefDetect: From Supervised Encoders to Hierarchical LLM-RAG for Psychological Defense Detection

DAL Team at PsyDefDetect: From Supervised Encoders to Hierarchical LLM-RAG for Psychological Defense Detection

Paper · 2026 · Duc-Luong Tran, Phuong-Anh Chu, Hoang-Dat Do et al.

Summary

The paper describes the DAL Team's system for the PsyDefDetect shared task at BioNLP 2026. The task is to assign one of nine psychological defense mechanisms, based on the DMRS taxonomy, to turns in multi-turn emotional-support dialogues from the PSYDEFCONV corpus. Two features make it hard: defenses are implicit and depend on context, and the labels are heavily skewed toward a few frequent classes.

The authors compare four approaches:

  • supervised RoBERTa–LSTM dialogue encoders with flat or cascaded prediction;
  • a retrieval-augmented pipeline built on Gemini 3.1 Flash Lite;
  • a hybrid that filters frequent labels with the supervised model before passing the rest to the LLM;
  • a RoBERTa classifier trained on Phi-4 context summaries.

The main system rebuilds the dialogue context and retrieves similar labeled examples, each paired with an LLM-written rationale. It then predicts a coarse defense group before choosing the specific label, and rule-based discourse cues adjust the final output. This hierarchical LLM-RAG variant gives the best macro-F1. It mainly helps rare classes, at some cost to overall accuracy.

The work matters because it shows both what LLMs can do and where they fall short when modeling latent clinical constructs in conversation.

Key points

  • Data imbalance: label 7 (mature defenses) covers about 52% of the training data and label 0 about 16%. In the test set, label 7 accounts for 243 of 397 samples.
  • Retrieval design: examples are matched symbolically rather than with neural embeddings, using Jaccard overlap and 17 rule-extracted discourse cues. The pipeline selects k = 3 examples with MMR (λ = 0.20) and allows at most 6 examples per prompt.
  • Coarse groups: the hierarchical step first sorts turns into No Defense (labels 0 and 8), Mature (7), Neurotic (5–6) or Immature (1–4).
  • Test results:
    • Hierarchical LLM-RAG: macro-F1 31.13%, precision 41.87%, accuracy 48.31%.
    • Direct LLM-RAG: macro-F1 26.55%.
    • 3-level supervised cascade: macro-F1 26.49%.
    • 2-level supervised cascade: best accuracy, 61.23%.
    • Hybrid pipeline: highest recall (30.19%), but lower F1 than the hierarchical system.
  • Trade-off and ranking: the official submission placed 10th of 21 teams. Under the hierarchical reformulation, class 8 F1 rose from 0 to roughly 60%, while class 7 dropped by about 12 points.

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

🎓VNU University of Engineering and TechnologyHanoi · Vietnam4 works · 1 citation

 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 worksPsychology · 1,552 worksSocial Psychology · 306 works

Cite

Cite as Duc-Luong Tran, Phuong-Anh Chu, Hoang-Dat Do, Tu-Phuong Mai, Duy-Cat Can and Hoang-Quynh Le. 2026. DAL Team at PsyDefDetect: From Supervised Encoders to Hierarchical LLM-RAG for Psychological Defense Detection. Proceedings of the BioNLP 2026 (Shared Tasks).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.bionlp-2.23

@inproceedings{tran2026dal,
  title = {{DAL Team at PsyDefDetect: From Supervised Encoders to Hierarchical LLM-RAG for Psychological Defense Detection}},
  author = {Duc-Luong Tran, Phuong-Anh Chu, Hoang-Dat Do, Tu-Phuong Mai, Duy-Cat Can and Hoang-Quynh Le},
  booktitle = {Proceedings of the BioNLP 2026 (Shared Tasks)},
  year = {2026},
  doi = {10.18653/v1/2026.bionlp-2.23},
  url = {https://doi.org/10.18653/v1/2026.bionlp-2.23}
}

References

24 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Eliseo Bao, Anxo Pérez, and Javier Parapar. 2024. Explainable depression symptom detection in social media. Health Information Science and Systems, 12(1):47. find
  2. Emily M Bender and Alexander Koller. 2020. Climbing towards nlu: On meaning, form, and understanding in the age of data. In Proceedings of the 58th annual meeting of the association for computational linguistics, pages 5185–5198.
  3. Jaime Carbonell and Jade Goldstein. 1998. The use of mmr, diversity-based reranking for reordering documents and producing summaries. In Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval, pages 335–336. find
  4. Stevie Chancellor and Munmun De Choudhury. 2020. Methods in predictive techniques for mental health status on social media: a critical review. NPJ digital medicine, 3(1):43. find
  5. Zhiyu Chen, Yujie Lu, and William Wang. 2023. Empowering psychotherapy with large language models: Cognitive distortion detection through diagnosis of thought prompting. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 4295–4304. find
  6. Mariagrazia Di Giuseppe, Katie Aafjes-van Doorn, Vera Békés, Bernard S Gorman, Karl Stukenberg, and Sherwood Waldron. 2024. Therapists’ defense use impacts their patients’ defensive functioning: a systematic case study. Research in Psychotherapy: Psychopathology, Process, and Outcome, 27(2):797. find
  7. Mariagrazia Di Giuseppe and J. Christopher Perry. 2021. The hierarchy of defense mechanisms: Assessing defensive functioning with the defense mechanisms rating scales q-sort. Frontiers in Psychology, 12. find
  8. Isaac R. Galatzer-Levy, Daniel McDuff, Vivek Natarajan, Alan Karthikesalingam, and Matteo Malgaroli. 2023. The capability of large language models to measure psychiatric functioning. arXiv preprint arXiv:2308.01834. — [arXiv:2308.01834](https://arxiv.org/abs/2308.01834) find
  9. Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. 2024. Retrieval-augmented generation for large language models: A survey. Preprint, arXiv:2312.10997.
  10. Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2024. Lost in the middle: How language models use long contexts. Transactions of the association for computational linguistics, 12:157–173.
  11. Siyang Liu, Chujie Zheng, Orianna Demasi, Sahand Sabour, Yu Li, Zhou Yu, Yong Jiang, and Minlie Huang. 2021. Towards emotional support dialog systems. In Proceedings of the 59th annual meeting of the association for computational linguistics and the 11th international joint conference on natural lan…
  12. Mounica Maddela, Megan Ung, Jing Xu, Andrea Madotto, Heather Foran, and Y-Lan Boureau. 2023. Training models to generate, recognize, and reframe unhelpful thoughts. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13641– 13660. find
  13. Marta Mrozowicz-Wrońska. 2023. Defense mechanisms in affective disorders-the state of the art. Psychiatria Polska, 57(1):197–206. find
  14. Hongbin Na, Yining Hua, Zimu Wang, Tao Shen, Beibei Yu, Lilin Wang, Wei Wang, John Torous, and Ling Chen. 2025. A survey of large language models in psychotherapy: Current landscape and future directions. In Findings of the Association for Computational Linguistics: ACL 2025, pages 7362–7376, Vienna… find
  15. Hongbin Na, Zimu Wang, Zhaoming Chen, Yining Hua, Rena Gao, Kailai Yang, Ling Chen, Wei Wang, Shaoxiong Ji, John Torous, and Sophia Ananiadou. 2026a. Overview of the psydefdetect shared task at bionlp 2026: Detecting levels of psychological defense mechanisms in supportive conversations. In Proceedi…
  16. Hongbin Na, Zimu Wang, Zhaoming Chen, Peilin Zhou, Yining Hua, Grace Ziqi Zhou, Haiyang Zhang, Tao Shen, Wei Wang, John Torous, Shaoxiong Ji, and Ling Chen. 2026b. You never know a person, you only know their defenses: Detecting levels of psychological defense mechanisms in supportive conversations.…
  17. J Christopher Perry. 2014. Anomalies and specific functions in the clinical identification of defense mechanisms. Journal of clinical psychology, 70(5):406– 418. find
  18. Hannah Rashkin, Eric Michael Smith, Margaret Li, and Y-Lan Boureau. 2019. Towards empathetic open-domain conversation models: A new benchmark and dataset. In Proceedings of the 57th annual meeting of the association for computational linguistics, pages 5370–5381. find
  19. Ashish Sharma, Adam Miner, David Atkins, and Tim Althoff. 2020. A computational approach to understanding empathy expressed in text-based mental health support. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5263–5276.
  20. Vishal Anand Shetty, Shauna Durbin, Meghan S Weyrich, Airín Denise Martínez, Jing Qian, and David L Chin. 2024. A scoping review of empathy recognition in text using natural language processing. Journal of the American Medical Informatics Association, 31(3):762–775. find
  21. Sichang Tu, Abigail Powers, Natalie Merrill, Negar Fani, Sierra Carter, Stephen Doogan, and Jinho D Choi. 2024. Automating ptsd diagnostics in clinical interviews: Leveraging large language models for trauma assessments. In Proceedings of the 25th Annual Meeting of the Special Interest Group on Disc… find
  22. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, and Denny Zhou. 2022. Chain-of-thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems, volume 35, pages 24824–24837.
  23. Xuhai Xu, Bingsheng Yao, Yuanzhe Dong, Saadia Gabriel, Hong Yu, James Hendler, Marzyeh Ghassemi, Anind K Dey, and Dakuo Wang. 2024. Mental-llm: Leveraging large language models for mental health prediction via online text data. Proceedings of the ACM on interactive, mobile, wearable and ubiquitous t… find
  24. Kailai Yang, Shaoxiong Ji, Tianlin Zhang, Qianqian Xie, Ziyan Kuang, and Sophia Ananiadou. 2023. Towards interpretable mental health analysis with large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 6056–6077. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tran-2026-dal-team-psydefdetect-supervised-encoders/