SubmissionNumber#=%=#29 FinalPaperTitle#=%=#CS_Metro at PsyDefDetect: Detecting Psychological Defense Mechanisms in Mental Health Dialogues with Summarization-Enhanced Transformer Ensembles ShortPaperTitle#=%=# NumberOfPages#=%=# CopyrightSigned#=%=# JobTitle#==# Organization#==# Abstract#==#Detecting psychological defense mechanisms in supportive conversations is essential for assisting mental health practitioners in monitoring client progress and tailoring therapeutic interventions. Natural language processing techniques are increasingly integral to such systems, enabling automated classification of defense levels to better understand help-seeker behavior and resistance patterns. In PsyDefDetect at BioNLP 2026, we address the task of nine-class defense level classification on the PSYDEFCONV corpus. We propose a three-stage pipeline combining LLM-based dialogue summarization, domain-specific transformer fine-tuning, and rule-based ensemble prediction. Additionally, we evaluate three mental health domain-specific transformers (Mental-BERT, Mental-RoBERTa, Mental-XLNet) alongside fine-tuned LLMs (Qwen3-4B, Qwen3-1.7B, Mistral-7B under different input conditions. Experimental results show that our ensemble approach achieves the best performance, reaching 34.69\% macro F1 and surpassing the baseline by 4.69 percentage points. These findings demonstrate that domain-specific transformers substantially outperform generic LLM fine-tuning on this specialized clinical task. Author{1}{Firstname}#=%=#Oarisa Author{1}{Lastname}#=%=#Rebayet Author{1}{Username}#=%=#oarisa_r Author{1}{Orcid}#=%=# Author{1}{Email}#=%=#orebayet@gmail.com Author{1}{Affiliation}#=%=#Chittagong University of Engineering and Technology Author{2}{Firstname}#=%=#Radiul Author{2}{Lastname}#=%=#Walee Author{2}{Username}#=%=#walee94 Author{2}{Orcid}#=%=# Author{2}{Email}#=%=#walee2radid151@gmail.com Author{2}{Affiliation}#=%=#Chittagong University of Engineering and Technology Author{3}{Firstname}#=%=#Symom Hossain Author{3}{Lastname}#=%=#Shohan Author{3}{Username}#=%=#symom_hossain_shohan Author{3}{Orcid}#=%=# Author{3}{Email}#=%=#u1904048@student.cuet.ac.bd Author{3}{Affiliation}#=%=#Chittagong University of Engineering and Technology Author{4}{Firstname}#=%=#Kawsar Author{4}{Lastname}#=%=#Ahmed Author{4}{Username}#=%=#kawsar Author{4}{Orcid}#=%=#https://orcid.org/0009-0002-7141-6196 Author{4}{Email}#=%=#u1804017@student.cuet.ac.bd Author{4}{Affiliation}#=%=#Chittagong University of Engineering and Technology Author{5}{Firstname}#=%=#Mohammed Moshiul Author{5}{Lastname}#=%=#Hoque Author{5}{Username}#=%=#moshiul Author{5}{Orcid}#=%=# Author{5}{Email}#=%=#moshiul_240@cuet.ac.bd Author{5}{Affiliation}#=%=#Department of Computer Science & Engineering, Chittagong University of Engineering & Technology ========== èéáğö