Core Concepts
Enhancing psychological support through large language models tailored for Cognitive Behavioral Therapy techniques.
Abstract
Introduction
AI advancements in mental health support.
Challenges in data quality and scarcity.
Introduction of CBT-LLM for structured responses.
Related Work
Overview of counseling techniques like CBT, ACT, DBT.
LLMs' role in mental health support.
Methodology
Problem definition using PsyQA dataset.
Generation of CBT responses based on prompts.
Data Extraction
"The recent advancements in artificial intelligence highlight the potential of language models in psychological health support."
"Empirical evaluations demonstrate that CBT-LLM excels in generating structured, professional, and highly relevant responses."
Quotations
"For many people socializing is an important need and for you it has been a year since you have socialized and this can make you feel very depressed and lonely."
Experiment
Data preparation and baselines for evaluation.
Main Results
CBT-LLM outperforms benchmark models like LLaMA-Chinese-7B and Alpaca-Chinese-7B.
Human Evaluation
Baichuan-7B marginally outperforms Alpaca-Chinese-7B in relevance, structure adherence, and helpfulness measures.
Case Study
Example response addressing social isolation concerns using CBT principles.
Conclusions and Future Work
Bridging LLMs with CBT for enhanced mental health support.
Limitations
Lack of well-defined annotations for cognitive distortions may impact response accuracy.
Ethical Statement
Release of datasets for research purposes only to protect privacy.
Acknowledgements
Recognition to evaluators, contributors, and dataset providers.
Stats
"The recent advancements in artificial intelligence highlight the potential of language models in psychological health support."
"Empirical evaluations demonstrate that CBT-LLM excels in generating structured, professional, and highly relevant responses."
Quotes
"For many people socializing is an important need and for you it has been a year since you have socialized and this can make you feel very depressed and lonely."