@inproceedings{tran-etal-2026-dal,
    title = "{DAL} Team at {P}sy{D}ef{D}etect: From Supervised Encoders to Hierarchical {LLM}-{RAG} for Psychological Defense Detection",
    author = "Tran, Duc-Luong  and
      Chu, Phuong-Anh  and
      Do, Hoang-Dat  and
      Mai, Tu-Phuong  and
      Can, Duy-Cat  and
      Le, Hoang-Quynh",
    editor = "Gupta, Deepak  and
      Demner-Fushman, Dina",
    booktitle = "Proceedings of the {B}io{NLP} 2026 (Shared Tasks)",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/corrections-2026-07/2026.bionlp-2.23/",
    doi = "10.18653/v1/2026.bionlp-2.23",
    pages = "164--170",
    ISBN = "979-8-89176-435-4",
    abstract = "The PsyDefDetect shared task focuses on classifying nine psychological defense mechanisms in multi-turn dialogues, a problem complicated by severe label imbalance and the implicit, context-dependent nature of defenses. In this work, we investigate several approaches for dialogue-level defense detection, including supervised baselines and large language model (LLM)-based pipelines. Our primary system is a retrieval-augmented LLM framework with hierarchical prediction and lightweight heuristics for decision calibration. Experiments on the PSYDEFCONV dataset show that LLM-based methods improve overall performance compared to supervised baselines, but still struggle with fine-grained distinctions, especially for minority labels. These findings highlight the challenges of modeling implicit psychological constructs in dialogue."
}