Personalized Empathy on PersonaEmp (Random Split)
4.53Qwen3 Resonation ScorePereGRM
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PereGRMTraining Strategy=Training-based, Evaluation Rounds (K)=82026.05 | 4.53 | 4.5 | 4.55 | 4.53 | 3.32 | 3.62 | 3.81 | 3.58 | |
| PereGRMTraining Strategy=Training-based, Evaluation Rounds (K)=12026.05 | 4.47 | 4.37 | 4.31 | 4.38 | 3.3 | 3.5 | 3.7 | 3.5 | |
| RLPATraining Strategy=Training-based2026.05 | 4.43 | 4.35 | 4.44 | 4.41 | 3.26 | 3.41 | 3.61 | 3.43 | |
| PERMTraining Strategy=Training-based2026.05 | 4.32 | 4.15 | 4.25 | 4.24 | 2.94 | 3.14 | 3.25 | 3.11 | |
| Qwen3-8BTraining Strategy=Base Model2026.05 | 3.7 | 3.74 | 3.76 | 3.73 | 2.62 | 2.95 | 3.02 | 2.86 | |
| MemoryTraining Strategy=Training-free2026.05 | 3.66 | 3.72 | 3.71 | 3.69 | 2.47 | 2.87 | 2.86 | 2.73 | |
| SFTTraining Strategy=Training-based2026.05 | 3.55 | 3.68 | 3.65 | 3.63 | 2.7 | 3.07 | 3.11 | 2.96 | |
| RAGTraining Strategy=Training-free2026.05 | 3.53 | 3.64 | 3.73 | 3.63 | 2.52 | 2.9 | 2.99 | 2.8 |