๐Ÿง  AI Mental Health Guard (MPNet Smart)

This model is a fine-tuned version of microsoft/mpnet-base specialized in detecting mental health statuses.

Capabilities

It can classify text into three categories:

  1. Depression
  2. Anxiety
  3. Suicide Risk (High Recall)

Special Training

Unlike standard models, this model underwent Adversarial Training to distinguish between:

  • Clinical Depression vs. Sadness over TV Shows/Movies
  • Suicide Ideation vs. Metaphorical speech (e.g., "dying of laughter")
  • Real Anger vs. Gaming rage

Performance

  • F1 Macro: ~0.79
  • Accuracy on Trick Cases: 100% (in tests)
Downloads last month
26
Safetensors
Model size
0.1B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Space using ekenger/mpnet-multilabel-mental-health 1