Huysun29/cbt-qwen2.5-7b-v2
Huysun29/cbt-qwen2.5-7b-v2 is a 7.6 billion parameter Qwen2.5-7B-Instruct fine-tune, specifically optimized for Cognitive Behavioral Therapy (CBT) consultation assistance. This model is designed to generate structured JSON outputs, including fields like risk_level, technique, and rationale, for therapeutic dialogue. Its primary strength lies in providing structured responses for CBT-related interactions, prioritizing safety recommendations in crisis situations.
Loading preview...
Huysun29/cbt-qwen2.5-7b-v2: CBT Consultation Assistant
This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct, specifically adapted for Cognitive Behavioral Therapy (CBT) consultation. It stands out by generating structured JSON outputs for therapeutic interactions, encompassing fields such as risk_level, technique, rationale, plan, response, and safety_action.
Key Capabilities
- Structured JSON Output: Provides consistent, machine-readable responses for CBT dialogues.
- CBT-Specific Fine-tuning: Optimized for therapeutic consultation, drawing from 26,000 CBT conversation samples.
- Safety-Oriented: Trained to prioritize safety and recommend professional help or emergency services in crisis or self-harm scenarios.
Training Details
The model was fine-tuned using LoRA (r=32, alpha=64) with a learning rate of 5e-05 and a maximum sequence length of 1536, achieving an evaluation loss of 0.4938511252403259.
Important Considerations
- Academic Support Tool: Intended as an academic aid, not a replacement for mental health professionals.
- No Diagnosis or Medication Advice: The model does not diagnose conditions or provide medication recommendations.
- Human-in-the-Loop Required: Professional review is essential before deployment in real-world therapeutic settings.