Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering

Hongbin Na

TL;DR

The paper presents CBT-LLM, a Chinese CBT-grounded QA model built by crafting a CBT-focused prompt and a CBT QA dataset (derived from PsyQA) to guide instruction-tuned fine-tuning with LoRA. Empirical results from automatic metrics and human evaluators indicate CBT-LLM delivers structured, professional, and CBT-consistent responses for psychological health support, outperforming several baselines. The work releases the CBT QA dataset and the CBT-LLM on Hugging Face, highlighting practical implications for CBT-based digital mental health assistance while acknowledging limitations such as annotation gaps and single-turn interaction. Future work aims to broaden therapeutic coverage (ACT/DBT) and move toward multi-turn dialogues to better emulate real counseling sessions.

Abstract

The recent advancements in artificial intelligence highlight the potential of language models in psychological health support. While models trained on data from mental health service platform have achieved preliminary success, challenges persist in areas such as data scarcity, quality, and ensuring a solid foundation in psychological techniques. To address these challenges, this study introduces a novel approach to enhance the precision and efficacy of psychological support through large language models. Specifically, we design a specific prompt derived from principles of Cognitive Behavioral Therapy (CBT) and have generated the CBT QA dataset, specifically for Chinese psychological health Q&A based on CBT structured intervention strategies. Unlike previous methods, our dataset emphasizes professional and structured response. Utilizing this dataset, we fine-tuned the large language model, giving birth to CBT-LLM, the large-scale language model specifically designed for Cognitive Behavioral Therapy techniques. Empirical evaluations demonstrate that CBT-LLM excels in generating structured, professional, and highly relevant responses in psychological health support tasks, showcasing its practicality and quality. The model is available on Hugging Face: https://huggingface.co/Hongbin37/CBT-LLM.

CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering

TL;DR

The paper presents CBT-LLM, a Chinese CBT-grounded QA model built by crafting a CBT-focused prompt and a CBT QA dataset (derived from PsyQA) to guide instruction-tuned fine-tuning with LoRA. Empirical results from automatic metrics and human evaluators indicate CBT-LLM delivers structured, professional, and CBT-consistent responses for psychological health support, outperforming several baselines. The work releases the CBT QA dataset and the CBT-LLM on Hugging Face, highlighting practical implications for CBT-based digital mental health assistance while acknowledging limitations such as annotation gaps and single-turn interaction. Future work aims to broaden therapeutic coverage (ACT/DBT) and move toward multi-turn dialogues to better emulate real counseling sessions.

Abstract

The recent advancements in artificial intelligence highlight the potential of language models in psychological health support. While models trained on data from mental health service platform have achieved preliminary success, challenges persist in areas such as data scarcity, quality, and ensuring a solid foundation in psychological techniques. To address these challenges, this study introduces a novel approach to enhance the precision and efficacy of psychological support through large language models. Specifically, we design a specific prompt derived from principles of Cognitive Behavioral Therapy (CBT) and have generated the CBT QA dataset, specifically for Chinese psychological health Q&A based on CBT structured intervention strategies. Unlike previous methods, our dataset emphasizes professional and structured response. Utilizing this dataset, we fine-tuned the large language model, giving birth to CBT-LLM, the large-scale language model specifically designed for Cognitive Behavioral Therapy techniques. Empirical evaluations demonstrate that CBT-LLM excels in generating structured, professional, and highly relevant responses in psychological health support tasks, showcasing its practicality and quality. The model is available on Hugging Face: https://huggingface.co/Hongbin37/CBT-LLM.
Paper Structure (23 sections, 3 equations, 4 figures, 5 tables)

This paper contains 23 sections, 3 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Counseling Techniques
  4. LLMs for Mental Health Support
  5. Methodology
  6. Problem Definition
  7. Generation of CBT Responses
  8. CBT Response Analysis
  9. General Statistical Analysis
  10. Cognitive Distortion Statistical Analysis
  11. Recognition of Quality Analysis
  12. CBT-LLM
  13. Experiment
  14. Data Preparation
  15. Baselines
  16. ...and 8 more sections

Figures (4)

  • Figure 1: An example of poor data quality. "Human" represents actual human responses from an online mental health forum, while "ChatGPT" denotes our responses generated based on Cognitive Behavioral Therapy (CBT) prompt.
  • Figure 2: An overview of training CBT-LLM. It first utilizes PsyQA Questions and CBT Prompt to generate CBT answers, and then fine-tuning CBT-LLM.
  • Figure 3: CBT prompt for dataset generation.
  • Figure 4: Five primary sections are distinctly highlighted using different colors: 1. Expression of validation and empathy (Green), 2. Identification of Key thought or beliefs (Yellow), 3. Introduction of challenges or reflections (Purple), 4. Provision of strategies or insights (Blue), and 5. Encouragement and foresight (Red).