Personalized Empathy on PersonaEmp (OOD)
4.57Qwen3 ResonationPereGRM
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PereGRMTraining Strategy=Training-based, Evaluation Rounds (K)=82026.05 | 4.57 | 4.28 | 4.48 | 4.44 | 3.23 | 3.33 | 3.57 | 3.38 | |
| PereGRMTraining Strategy=Training-based, Evaluation Rounds (K)=12026.05 | 4.43 | 4.28 | 4.31 | 4.34 | 3.21 | 3.29 | 3.46 | 3.32 | |
| PERMTraining Strategy=Training-based2026.05 | 4.26 | 3.88 | 3.99 | 4.04 | 3.18 | 3.2 | 3.27 | 3.22 | |
| RLPATraining Strategy=Training-based2026.05 | 4.24 | 4.1 | 4.23 | 4.19 | 3.09 | 3.28 | 3.46 | 3.28 | |
| Qwen3-8BTraining Strategy=Base Model2026.05 | 3.61 | 3.63 | 3.67 | 3.64 | 2.6 | 2.91 | 2.96 | 2.82 | |
| SFTTraining Strategy=Training-based2026.05 | 3.53 | 3.64 | 3.57 | 3.58 | 2.73 | 3.07 | 3.05 | 2.95 | |
| MemoryTraining Strategy=Training-free2026.05 | 3.5 | 3.65 | 3.59 | 3.58 | 2.42 | 2.83 | 2.89 | 2.71 | |
| RAGTraining Strategy=Training-free2026.05 | 3.45 | 3.51 | 3.6 | 3.52 | 2.5 | 2.93 | 2.92 | 2.78 |