SoulChat 2.0 - A digital twin language model for psychological counselors launched by South China University of Technology.

What is SoulChat 2.0?

SoulChat 2.0 is a large-scale digital twin language model for psychological counselors, developed by the Future Technology Institute of South China University of Technology - Guangdong Provincial Key Laboratory of Digital Twins, based on the SoulChat 1.0 model. It defines specific digital twin tasks for psychological counselors for the first time, aiming to improve the performance of the large model in real psychological counseling scenarios by simulating the language style and therapeutic techniques of real psychological counselors. SoulChat 2.0 can build large-scale mental health models with specific psychological counselor styles at low cost, quickly, and efficiently, assisting psychological counselors in their work, such as pre-conversation interviews and providing 24-hour online services.

Main features of SoulChat 2.0

• Digital twin modeling of psychological counselorsSoulChat 2.0 can achieve digital twins of specific psychological counselors through fine-tuning, simulating the counselor's language style and counseling techniques, thereby improving the application performance of LLMs in real psychological counseling scenarios.
• Low-cost, fast, and efficient model buildingSoulChat 2.0 can use a small number of real counseling cases to create digital twin models of psychological counselors, enabling the construction of large-scale mental health models with specific counselor styles in a low-cost, fast, and efficient manner.
• Supporting psychological counseling workSoulChat 2.0 can assist real-world psychotherapists in their work, such as conducting pre-conference interviews and providing 24-hour online services.
• Open source collaborationThe project methodology, datasets, and models of SoulChat 2.0 have been fully open-sourced, facilitating comparison and reproduction by the research community and promoting the development of large-scale mental health models.

The technical principles of SoulChat 2.0

• Digital twin data generation for psychological counselorsSoulChat 2.0 generates digital twin data for single-turn conversations of psychotherapists by integrating the language style, counseling techniques, and client Big Five personality traits of real-world counselors with real-world counseling cases. This process can quickly generate batches of counseling cases for digital twin modeling of that psychotherapist.
• Multi-turn dialogue synthesisSoulChat 2.0 uses large-scale language models (LLMs) to build digital twins of psychotherapists with personalized counseling styles. This involves dynamic one-time learning from counseling cases, client-side personality simulation, and multi-round synthesis of mental health dialogues.
• Fine-tuning large language modelsGiven counseling case data for modeling a digital twin of a therapist, SoulChat 2.0 can fine-tune the process to create a digital twin of that therapist. This involves using a specific base model and performing full fine-tuning on the PsyDTCorpus training set to create a digital twin of a therapist with a specific counseling style.
• Multi-dimensional assessment and improvementSoulChat 2.0 has made significant improvements over other models in terms of conversation techniques, state and attitude, emotional empathy, and cognitive empathy, thus enhancing the model's psychological counseling performance.

The project address for SoulChat 2.0

• Github repository:https://github.com/scutcyr/SoulChat2.0

Application Scenarios of SoulChat 2.0

• Psychological counseling assistanceSoulChat 2.0 can simulate the language style and therapeutic techniques of real-world psychotherapists, assisting them in their work. It can be used for pre-conference interviews and provides 24/7 online service to meet the needs of users at different times.
• Large Model Study of Mental HealthSoulChat 2.0 uses a small number of real consultation cases to create digital twin models of psychological counselors, bringing new research ideas to the field of large-scale mental health models.