Doan Nam Long Vu | Simone Balloccu

Position: PhD Student

Topics: Multimodal AI for mental health

Bio

I am a researcher working on NLP, large language models, and clinical AI, with a focus on interpretability and mental health applications. My research investigates how language models encode and respond to clinically relevant content, particularly how the register in which symptoms are described shapes model behavior and assessments. A central question driving my work is whether LLMs, when applied to mental health contexts, are responding to underlying clinical conditions or to the surface-level language used to express them.

Additionally, I am exploring the deployment of state-of-the-art, open-source multimodal AI models in real-world clinical research scenarios. My applied work includes transforming clinical questionnaires into natural dialogues, developing dialogue-aware text-to-speech system, and utilizing automatic speech recognition for niche use cases, such as analyzing interactions between caretakers and children with autism. Furthermore, I am experimenting with video-based models for gaze tracking and behavioral classification in children.

Research Interests

Natural Language Processing
Large Language Models
Multimodal AI
Model Interpretability
Clinical AI & Mental Health Applications
Text-to-Speech & Automatic Speech Recognition

Publications

2026

Prompts Without Evidence: How Neuroimaging Mentions Shift Clinical Vision-Language Model Predictions

Doan Nam Long Vu and Simone Balloccu

arXiv preprint arXiv:2603.28387 , 2026

Accepted at EMNLP 2026 Main, pending publication.

Abs Paper

Trustworthy clinical AI must use real evidence and avoid relying on surface-level artifacts. We evaluate 12 open-weight vision-language models (VLMs) on two clinical neuroimaging cohorts for binary classification of affective disorders and cognitive decline. Both cohorts include structural magnetic resonance imaging (MRI) acquired under their original research protocols. Prior work does not establish the included neuroimaging inputs as reliable stand-alone diagnostic evidence for the present tasks. Nevertheless, when neuroimaging context is introduced, smaller VLMs gain up to 0.66 F1 under the evaluated augmented conditions, becoming competitive with models an order of magnitude larger. Confidence estimation shows that most of the calibration improvement for the analyzed smaller models occurs after the MRI reference is added to the prompt, before any image is supplied. Our preliminary expert case study finds that faithfulness remains low in every condition examined, with the reviewed model introducing unverified clinical details. Finally, in our single-model intervention, preference alignment suppresses MRI-referencing behavior but reduces the augmented-condition advantage, leaving the underlying issue unresolved. These results caution against reading surface metric gains as evidence of true multimodal integration, with direct implications for clinical VLM deployment.

2025

Roleplaying with Structure: Synthetic Therapist-Client Conversation Generation from Questionnaires

Doan Nam Long Vu, Rui Tan, Lena Moench, and 8 more authors

arXiv preprint arXiv:2510.25384 , 2025

Accepted at NLP4PI Workshop 2026, pending publication.

Abs Paper

The development of AI for mental health is hindered by a lack of authentic therapy dialogues, due to strict privacy regulations and the fact that clinical sessions were historically rarely recorded. We present an LLM-driven pipeline that generates synthetic counseling dialogues based on structured client profiles and psychological questionnaires. Grounded on the principles of Cognitive Behavioral Therapy (CBT), our method creates synthetic therapeutic conversations for clinical disorders such as anxiety and depression. Our framework, SQPsych (Structured Questionnaire-based Psychotherapy), converts structured psychological input into natural language dialogues through therapist-client simulations. Due to data governance policies and privacy restrictions prohibiting the transmission of clinical questionnaire data to third-party services, previous methodologies relying on proprietary models are infeasible in our setting. We address this limitation by generating a high-quality corpus using open-weight LLMs, validated through human expert evaluation and LLM-based assessments. Our SQPsychLLM models fine-tuned on SQPsychConv achieve strong performance on counseling benchmarks, surpassing baselines in key therapeutic skills. Our findings highlight the potential of synthetic data to enable scalable, data-secure, and clinically informed AI for mental health support. We will release our code, models, and corpus at this https URL.