Output Diversity on MT-Bench
48.36Lexical Diversity ScoreQEMPO-KL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| QEMPO-KLBase Model=Llama-3.2-1B-Instruct2026.02 | 48.36 | 30.55 | 53.38 | 44.1 | |
| SPLBase Model=Llama-3.2-1B-Instruct2026.02 | 47.88 | 30.83 | 52.18 | 43.63 | |
| QEMPOBase Model=Llama-3.2-1B-Instruct2026.02 | 46.51 | 30.84 | 53.95 | 43.77 | |
| QEMPOBase Model=Qwen2.5-1.5B-Instruct2026.02 | 44.23 | 30.18 | 54.29 | 42.9 | |
| QEMPO-KLBase Model=Qwen2.5-1.5B-Instruct2026.02 | 44.18 | 30.08 | 53.3 | 42.52 | |
| RLHFBase Model=Llama-3.2-1B-Instruct2026.02 | 43.85 | 29.08 | 52.43 | 41.79 | |
| SPLBase Model=Qwen2.5-1.5B-Instruct2026.02 | 43.53 | 30.09 | 54.52 | 42.71 | |
| Qwen2.5-1.5B-InstructBase Model=Qwen2.5-1.5B-Instruct2026.02 | 41.27 | 29.2 | 53.39 | 41.29 | |
| Llama-3.2-1B-InstructBase Model=Llama-3.2-1B-Instruct2026.02 | 40.78 | 29.54 | 53.55 | 41.29 | |
| RLHFBase Model=Qwen2.5-1.5B-Instruct2026.02 | 40.42 | 28.74 | 51.92 | 40.36 | |
| SPLBase Model=Qwen2.5-7B-Instruct2026.02 | 36.18 | 30.31 | 55.77 | 40.75 | |
| QEMPOBase Model=Llama-3.1-8B-Instruct2026.02 | 35.6 | 28.7 | 53.42 | 39.24 | |
| QEMPOBase Model=Qwen2.5-7B-Instruct2026.02 | 35.37 | 28.34 | 52.45 | 38.72 | |
| Llama-3.1-8B-InstructBase Model=Llama-3.1-8B-Instruct2026.02 | 34.85 | 29.55 | 54.12 | 39.51 | |
| QEMPO-KLBase Model=Qwen2.5-7B-Instruct2026.02 | 34.6 | 28.08 | 52.34 | 38.34 | |
| RLHFBase Model=Llama-3.1-8B-Instruct2026.02 | 34.5 | 27.73 | 51.85 | 38.03 | |
| QEMPO-KLBase Model=Llama-3.1-8B-Instruct2026.02 | 33.63 | 30.68 | 54.91 | 39.74 | |
| RLHFBase Model=Qwen2.5-7B-Instruct2026.02 | 31.98 | 27.41 | 51.48 | 36.95 | |
| SPLBase Model=Llama-3.1-8B-Instruct2026.02 | 31.81 | 27.41 | 52.34 | 37.19 | |
| Qwen2.5-7B-InstructBase Model=Qwen2.5-7B-Instruct2026.02 | 31.68 | 28.04 | 52.28 | 37.33 |