Personality-conditioned Text Generation on PERSONALITYBENCH (Human Evaluation test set)
2.33Average Rank (Agreeableness)Simple Prompt
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Simple PromptBase Model=LLaMA-3-8B-Instruct2024.10 | 2.33 | 2.63 | 2.58 | 3.15 | 3.25 | 2.79 | |
| NPTIBase Model=LLaMA-3-8B-Instruct2024.10 | 2.4 | 2.51 | 2.09 | 1.85 | 2.48 | 2.27 | |
| P2Base Model=LLaMA-3-8B-Instruct2024.10 | 2.41 | 2.41 | 2.39 | 2.28 | 2.13 | 2.32 | |
| SFTBase Model=LLaMA-3-8B-Instruct2024.10 | 2.45 | 2.49 | 2.21 | 2.6 | 2.12 | 2.37 | |
| PASBase Model=LLaMA-3-8B-Instruct2024.10 | 3.21 | 3.31 | 3.8 | 3.69 | 3 | 3.4 |