Patient Simulation on Client-Therapist Interactions (Bad Therapist persona)
4.9Factual ConsistencyPATIENT-Ψ
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PATIENT-ΨEvaluation Protocol=LLM judge, 5-point Likert scale2026.02 | 4.9 | 4.3 | 4.15 | 3.55 | 3.65 | 4.25 | 4 | 3.3 | 3.8 | 81.02 | 82.02 | 0.07 | |
| ClientCASTEvaluation Protocol=LLM judge, 5-point Likert scale2026.02 | 4.71 | 4.24 | 4.08 | 3.58 | 3.56 | 3.85 | 4.16 | 2.91 | 3.3 | 26.51 | 27.51 | 0.06 | |
| EeyoreEvaluation Protocol=LLM judge, 5-point Likert scale2026.02 | 4.68 | 4.03 | 4.12 | 3.48 | 3.42 | 3.82 | 4.12 | 2.72 | 3.28 | 21.48 | 23.17 | — | |
| AnnaAgentEvaluation Protocol=LLM judge, 5-point Likert scale2026.02 | 4.68 | 4.17 | 4.1 | 3.52 | 3.57 | 3.93 | 4.12 | 2.93 | 3.33 | 85.99 | 122.28 | 0.19 | |
| Roleplay-dohEvaluation Protocol=LLM judge, 5-point Likert scale2026.02 | 4.67 | 4.27 | 4.03 | 3.61 | 3.58 | 3.88 | 4.13 | 3.02 | 3.3 | 81.45 | 501.35 | 0.17 |