Therapeutic Dialogue Generation on Therapeutic Dialogue Dataset (test)
1.73Therapeutic RapportRL Model
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RL ModelBackbone=GPT-2, Training Mode=Reinforcement Learning (RL), Evaluation Protocol=LLM-as-a-judge2025.11 | 1.73 | 1.67 | 1.83 | 1.49 | 1.95 | 1.64 | |
| SFT ModelBackbone=GPT-2, Training Mode=Supervised Fine-Tuning (SFT), Evaluation Protocol=LLM-as-a-judge2025.11 | 1.7 | 1.57 | 1.74 | 1.48 | 1.91 | 1.53 | |
| Baseline GPT-2Backbone=GPT-2, Training Mode=Baseline, Evaluation Protocol=LLM-as-a-judge2025.11 | 1.09 | 1.03 | 1.07 | 1.02 | 1.1 | 1.06 |