aiqbench analysis
This research demonstrates the feasibility of using fine-tuned large language models to estimate depression severity from unstructured user dialogue, effectively addressing data scarcity through pseudo-labeling techniques. By achieving strong clinical metrics without relying on completed self-report questionnaires, the study highlights a viable path toward passive mental health monitoring in AI applications. The work underscores the potential for LLMs to extract continuous symptom records that traditional sparse datasets cannot capture.
Fine-tuning LLMs for Passive Depression Severity Estimation from AI Mental Health Dialogue
arXiv:2606.17973v2 Announce Type: replace Abstract: Depression is a leading cause of disability worldwide, yet validated questionnaires such as the Patient Health Questionnaire-9 (PHQ-9) often go uncompleted, yielding sparse, biased data. Users of AI mental health applications often describe their mood, sleep and functioning unprompted, producing a continuous record of symptoms. Here, we predict PHQ-9 scores from these conversations alone. As in many clinical language tasks, labels are scarce and imbalanced. We address both by fine-tuning Qwen 3.5 27B with a regression head and augmenting 3,111 ground-truth labels with pseudo-labels from a frontier reasoning model and iteratively trained intermediate models, expanding to 5,524 users with more balanced scores. On 422 held-out users, the model achieved MAE $= 3.98$, Pearson $r = 0.67$, and AUC $= 0.88$ at the clinical threshold (PHQ-9 $\geq 10$); AUC was $> 0.80$ across severity thresholds from PHQ-9 $\geq 3$ to PHQ-9 $\geq 24$, demonstrating that the model captures depression severity across the full clinical spectrum. These results support passive depression monitoring without self-report.
Abstract:Depression is a leading cause of disability worldwide, yet validated questionnaires such as the Patient Health Questionnaire-9 (PHQ-9) often go uncompleted, yielding sparse, biased data. Users of AI mental health applications often describe their mood, sleep and functioning unprompted, producing a continuous record of symptoms. Here, we predict PHQ-9 scores from these conversations alone. As in many clinical language tasks, labels are scarce and imbalanced. We address both by fine-tuning Qwen 3.5 27B with a regression head and augmenting 3,111 ground-truth labels with pseudo-labels from a frontier reasoning model and iteratively trained intermediate models, expanding to 5,524 users with more balanced scores. On 422 held-out users, the model achieved MAE $= 3.98$, Pearson $r = 0.67$, and AUC $= 0.88$ at the clinical threshold (PHQ-9 $\geq 10$); AUC was $> 0.80$ across severity thresholds from PHQ-9 $\geq 3$ to PHQ-9 $\geq 24$, demonstrating that the model captures depression severity across the full clinical spectrum. These results support passive depression monitoring without self-report.
Computation and Language (cs.CL)
arXiv:2606.17973 [cs.CL]
(or arXiv:2606.17973v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.17973 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 255 words
Related coverage
Open Distillation of Hereditary Traits
AI Alignment Forum · Jul 14, 2026
[AINews] OpenAI’s Patch the Planet program is 1 of 2 parties to discover PHP2Shell vulnerabilities in core PHP affecting 70% of the web, using GPT-5.6 Sol; coordinated response over 2 months
Latent Space · Oct 22, 2026
You Can Catch Sleeper Agents by Teaching Another Model to Imitate Them
LessWrong · Jun 10, 2026