Multi-turn Dialogue Evaluation on MT-Bench (MT-Bench Score, ŝu, ŝw)
8.99MT-Bench ScoreGPT-4
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-42026.01 | 8.99 | 0.83 | 0.73 | |
| GPT-3.52026.01 | 7.94 | 0.51 | 0.43 | |
| Vicuna-13B (all)#Token=370M2026.01 | 6.39 | -0.29 | -0.25 | |
| Alpaca-13B#Token=4.4M2026.01 | 4.53 | -0.62 | -0.54 | |
| LLaMA-13B#Token=1T2026.01 | 2.61 | -1.27 | -1.12 |