Interpretable Symptom Vectors for Depression in a Large Language Model

• URL: http://arxiv.org/abs/2609.01832v1
• Date: Tue, 01 Sep 2026 20:07:03 GMT
• Title: Interpretable Symptom Vectors for Depression in a Large Language Model
• Abstract summary: We analyzed the residual stream of Gemma-3-27B-PT using mechanistic interpretability techniques.<n>Recording activations across symptom descriptions drawn from validated clinical instruments, we found that symptom groups geometrically separated the most.<n>Using Semantic Projection, we then projected held-out naturalistic text onto Symptom Vectors constructed from these instruments.
• Score: 1.3546242205182983
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Patients with depression present with diverse symptom profiles, yet clinical practice routinely reduces this variation to a single severity score. Large language models (LLMs) can potentially capture various symptoms and their severity from patient speech. However, how depressive symptoms are represented inside LLMs remains poorly understood, limiting clinical trust. To examine whether internal model activations match clinician judgment, we analyzed the residual stream of Gemma-3-27B-PT using mechanistic interpretability techniques. Recording activations across symptom descriptions drawn from validated clinical instruments, we found that symptom groups geometrically separated the most at layer 21 across multiple distance metrics. Using Semantic Projection, we then projected held-out naturalistic text onto Symptom Vectors constructed from these instruments. The resulting per-symptom coefficients preserved clinician-annotated rank ordering across mood, somatic, and suicidality axes. Furthermore, a single depression vector in Layer 21 separates held-out depressive from non-depressive text (AUC = 0.789), which can be used as an emotional valence gate that restricts symptom projection to depressive speech. These results reveal a decorrelated, clinician-aligned symptom signal readable directly from internal activations, offering a mechanistic foundation for interpretable depression-assessment tools.

Related papers

• Psychosis involves a deficit of information compression in connected speech [4.907371324386136]
  Large language models (LLMs) with human-like performance on linguistic tasks have transformed the study of language in neurodiverse conditions.<n>Previous crosslinguistic evidence suggests a complexity reduction in the form of both lower intrinsic dimensionality (ID) of LLM representations and higher mean surprisal (prediction error) in psychosis.<n>We operationalized surprisal difference as the difference between surprisal as estimated from word frequency and surprisal as based on a contextual LM, which is sensitive to grammatical organization over and above lexical concepts.
  arXiv  Detail & Related papers  (2026-09-14T13:13:30Z)
• Depression Markers in Speech: An Approach based on Tract Variables Dynamics [52.72895760734978]
  This study identifies new depression biomarkers based on the dynamical properties of tract variables.<n>The proposed biomarkers effectively discriminate between the depressed and control speakers, as evidenced by the high Cliffs delta values across both read and spontaneous speech.
  arXiv  Detail & Related papers  (2026-07-28T15:48:07Z)
• Psychologically-Grounded Graph Modeling for Interpretable Depression Detection [36.72827968050813]
  We propose PsyGAT, a framework that models conversational sessions as dynamic temporal graphs.<n>We also introduce Causal-PsyGAT, an interpretability module that identifies symptom triggers.<n> Experiments show a 20% improvement in MRR for identifying causal indicators.
  arXiv  Detail & Related papers  (2026-04-27T07:27:45Z)
• Clinically Inspired Symptom-Guided Depression Detection from Emotion-Aware Speech Representations [6.043112894299487]
  Depression manifests through a diverse set of symptoms such as sleep disturbance, loss of interest, and concentration difficulties.<n>Most existing works treat depression prediction either as a binary label or an overall severity score without explicitly modeling symptom-specific information.<n>We propose a symptom-specific and clinically inspired framework for depression severity estimation from speech.
  arXiv  Detail & Related papers  (2026-02-17T13:47:05Z)
• DepFlow: Disentangled Speech Generation to Mitigate Semantic Bias in Depression Detection [54.209716321122194]
  We present DepFlow, a depression-conditioned text-to-speech framework.<n>A Depression Acoustic Camouflage learns speaker- and content-invariant depression embeddings through adversarial training.<n>A flow-matching TTS model with FiLM modulation injects these embeddings into synthesis, enabling control over depressive severity.<n>A prototype-based severity mapping mechanism provides smooth and interpretable manipulation across the depression continuum.
  arXiv  Detail & Related papers  (2026-01-01T10:44:38Z)
• Quantifying depressive mental states with large language models [0.0]
  Large Language Models (LLMs) may have an important role to play in mental health.<n>We outline and evaluate LLM performance on three critical tests.
  arXiv  Detail & Related papers  (2025-02-13T16:52:06Z)
• Hierarchical attention interpretation: an interpretable speech-level transformer for bi-modal depression detection [6.561362931802501]
  Depression is a common mental disorder. Automatic depression detection tools using speech, enabled by machine learning, help early screening of depression. This paper addresses two limitations that may hinder the clinical implementations of such tools: noise resulting from segment-level labelling and a lack of model interpretability.
  arXiv  Detail & Related papers  (2023-09-23T20:48:58Z)
• The Relationship Between Speech Features Changes When You Get Depressed: Feature Correlations for Improving Speed and Performance of Depression Detection [69.88072583383085]
  This work shows that depression changes the correlation between features extracted from speech. Using such an insight can improve the training speed and performance of depression detectors based on SVMs and LSTMs.
  arXiv  Detail & Related papers  (2023-07-06T09:54:35Z)
• Handwriting and Drawing for Depression Detection: A Preliminary Study [53.11777541341063]
  Short-term covid effects on mental health were a significant increase in anxiety and depressive symptoms. The aim of this study is to use a new tool, the online handwriting and drawing analysis, to discriminate between healthy individuals and depressed patients.
  arXiv  Detail & Related papers  (2023-02-05T22:33:49Z)
• Deep Multi-task Learning for Depression Detection and Prediction in Longitudinal Data [50.02223091927777]
  Depression is among the most prevalent mental disorders, affecting millions of people of all ages globally. Machine learning techniques have shown effective in enabling automated detection and prediction of depression for early intervention and treatment. We introduce a novel deep multi-task recurrent neural network to tackle this challenge, in which depression classification is jointly optimized with two auxiliary tasks.
  arXiv  Detail & Related papers  (2020-12-05T05:14:14Z)
• Detecting Parkinsonian Tremor from IMU Data Collected In-The-Wild using Deep Multiple-Instance Learning [59.74684475991192]
  Parkinson's Disease (PD) is a slowly evolving neuro-logical disease that affects about 1% of the population above 60 years old. PD symptoms include tremor, rigidity and braykinesia. We present a method for automatically identifying tremorous episodes related to PD, based on IMU signals captured via a smartphone device.
  arXiv  Detail & Related papers  (2020-05-06T09:02:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.