# TherapyGym: Evaluating and Aligning Clinical Fidelity and Safety in Therapy Chatbots

arXiv: https://arxiv.org/abs/2603.18008 · HTML: https://mlsanity.com/paper/2603.18008/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 51 references that are in the map.
- First posted on arXiv 23 Feb 2026.

## Builds on (most cited first, 22 of 22)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [JudgeBench: A Benchmark for Evaluating LLM-based Judges](https://mlsanity.com/paper/2410.12784/) (2024)
- [USR: An Unsupervised and Reference Free Evaluation Metric for Dialog Generation](https://mlsanity.com/paper/2005.00456/) (2020)
- [MT-Eval: A Multi-Turn Capabilities Evaluation Benchmark for Large Language Models](https://mlsanity.com/paper/2401.16745/) (2024)
- [GRADE: Automatic Graph-Enhanced Coherence Metric for Evaluating Open-Domain Dialogue Systems](https://mlsanity.com/paper/2010.03994/) (2020)
- [Towards Facilitating Empathic Conversations in Online Mental Health Support: A Reinforcement Learning Approach](https://mlsanity.com/paper/2101.07714/) (2021)
- [Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers](https://mlsanity.com/paper/2504.18412/) (2025)
- [Deconstruct to Reconstruct a Configurable Evaluation Metric for Open-Domain Dialogue Systems](https://mlsanity.com/paper/2011.00483/) (2020)
- [RLVER: Reinforcement Learning with Verifiable Emotion Rewards for Empathetic Agents](https://mlsanity.com/paper/2507.03112/) (2025)
- [PATIENT-Psi: Using Large Language Models to Simulate Patients for Training Mental Health Professionals](https://mlsanity.com/paper/2405.19660/) (2024)
- [ESC-Eval: Evaluating Emotion Support Conversations in Large Language Models](https://mlsanity.com/paper/2406.14952/) (2024)
- [PsyEval: A Suite of Mental Health Related Tasks for Evaluating Large Language Models](https://mlsanity.com/paper/2311.09189/) (2023)
- [CBT-Bench: Evaluating Large Language Models on Assisting Cognitive Behavior Therapy](https://mlsanity.com/paper/2410.13218/) (2024)
- [Preference Learning Unlocks LLMs' Psycho-Counseling Skills](https://mlsanity.com/paper/2502.19731/) (2025)
- [Psi-Arena: Interactive Assessment and Optimization of LLM-based Psychological Counselors with Tripartite Feedback](https://mlsanity.com/paper/2505.03293/) (2025)
- [ESC-Judge: A Framework for Comparing Emotional Support Conversational Agents](https://mlsanity.com/paper/2505.12531/) (2025)
- [TreeSynth: Synthesizing Diverse Data from Scratch via Tree-Guided Subspace Partitioning](https://mlsanity.com/paper/2503.17195/) (2025)
- [CPsyExam: A Chinese Benchmark for Evaluating Psychology using Examinations](https://mlsanity.com/paper/2405.10212/) (2024)

## Built on by (most cited first, 1 of 1)

- [Training Therapeutic Judges and Multi-Agent Systems for Human-Aligned Mental Health Support](https://mlsanity.com/paper/2606.30887/) (2026)

## Abstract (arXiv)

Large language models (LLMs) are increasingly used for mental-health support; yet prevailing evaluation methods--fluency metrics, preference tests, and generic dialogue benchmarks--fail to capture the clinically critical dimensions of psychotherapy. We introduce THERAPYGYM, a framework that evaluates and improves therapy chatbots along two clinical pillars: fidelity and safety. Fidelity is measured using the Cognitive Therapy Rating Scale (CTRS), implemented as an automated pipeline that scores adherence to CBT techniques over multi-turn sessions. Safety is assessed using a multi-label annotation scheme, covering therapy-specific risks (e.g., failing to address harm or abuse). To mitigate bias and unreliability in LLM-based judges, we further release THERAPYJUDGEBENCH, a validation set of 116 dialogues with 1,270 expert ratings for auditing and calibration against licensed clinicians. THERAPYGYM also serves as a training harness: CTRS and safety-based rewards drive RL with configurable patient simulations spanning diverse symptom profiles. Models trained in THERAPYGYM improve on expert ratings, with average CTRS rising from 0.10 to 0.60 (and 0.16 to 0.59 under LLM judges). Our work enables scalable development of therapy chatbots that are faithful to evidence-based practice and safer in high-stakes use.