LLM-powered psychotherapy simulator trains therapists at scale
Researchers have built MyMentorLLM, a multimodal simulation platform that uses LLMs to train psychotherapists at scale. The system generates realistic CBT sessions pairing trainee therapists with AI-driven patients exhibiting DSM-5-grounded disorders and expert supervisors, addressing a critical bottleneck in clinical education where human supervision capacity constrains training throughput. The 2,100-session dataset was evaluated on therapeutic competence and diagnostic accuracy, establishing a template for using generative AI to automate high-stakes professional credentialing workflows. This signals growing viability of LLMs in regulated domains requiring structured evaluation and expert feedback loops.
Modelwire context
ExplainerThe critical constraint MyMentorLLM addresses isn't just that therapist training is slow, but that human supervisors are the bottleneck. The platform's actual innovation is the three-way interaction model (trainee, AI patient, AI supervisor) that closes the feedback loop without requiring scarce expert time for every session.
This connects to the WorkSurface-Bench paper from the same day, which tackled knowledge routing in enterprise agents. Both papers solve a similar structural problem: systems that must route queries to the right resource before acting. Here, the system routes trainee decisions through expert-validated diagnostic frameworks (DSM-5) and competence rubrics, similar to how WorkSurface routes queries to the right knowledge surface. The difference is stakes: WorkSurface optimizes for accuracy in corporate settings, while MyMentorLLM must maintain fidelity in clinical credentialing where errors carry patient safety implications.
If MyMentorLLM's 2,100-session competence scores correlate with downstream clinical performance when trainees move to supervised human practice, that validates the simulation-to-real-world transfer. If they don't correlate, the platform is a practice tool, not a credentialing substitute. Watch whether any accrediting body (APA, BACP) formally recognizes hours logged on MyMentorLLM toward licensure requirements within 18 months.
Coverage we drew on
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsMyMentorLLM · Cognitive Behavioural Therapy · DSM-5-TR
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. arXiv cs.CL originally reported this story as “MyMentorLLM: A psychotherapy GenAI environment with multimodal voice/text patients, trainees and experts for deliberate practice”. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.