Multi-turn conversation on MT-Eval
8.28AccuracyRhea
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Rhea2025.12 | 8.28 | 14.42 | |
| Reply-Soft-Compress2025.12 | 7.83 | 15.21 | |
| BM25(RAG)2025.12 | 7.82 | 8.49 | |
| Vanilla2025.12 | 7.77 | 11.41 | |
| Recent-k2025.12 | 7.76 | 8.79 | |
| Summary2025.12 | 6.83 | 17.26 | |
| MemochaBase Model=Vicuna-7B2025.12 | 6.4 | 4.8 | |
| LongAlpacaBase Model=Vicuna-7B2025.12 | 5.89 | 5.44 | |
| LlmLingua22025.12 | 4.18 | 13.3 |