Persona-grounded dialogue generation on PersonaChat 200 random samples (test)
2.36Persona ConsistencyHyPE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HyPEBackbone=LLaMA-3.2-3B2026.06 | 2.36 | 2.56 | 3.34 | |
| MeanPoolBackbone=LLaMA-3.2-3B2026.06 | 2.27 | 2.45 | 3.31 | |
| MeanPoolBackbone=Qwen2.5-3B2026.06 | 2.26 | 2.64 | 3.34 | |
| HyPEBackbone=Qwen2.5-3B2026.06 | 2.26 | 2.58 | 3.4 | |
| HyPE-baseBackbone=Qwen2.5-3B2026.06 | 2.25 | 2.54 | 3.4 | |
| HyPE-baseBackbone=LLaMA-3.2-3B2026.06 | 2.19 | 2.56 | 3.22 | |
| HyPEBackbone=GPT-22026.06 | 1.74 | 1.82 | 2.41 | |
| HyPE-baseBackbone=GPT-22026.06 | 1.63 | 1.73 | 2.29 | |
| MeanPoolBackbone=GPT-22026.06 | 1.57 | 1.4 | 2.23 | |
| ORIGBackbone=GPT-22026.06 | 1.36 | 1.21 | 2.4 | |
| Text BaselineBackbone=GPT-22026.06 | 1.15 | 1.17 | 1.55 |