Dialogue Response Generation on KEEM (KMSC memories) 1.0 (test)
5.47PerplexityLlama10.8B tuned Korean
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Llama10.8B tuned Koreanparameters=10.8B, tuning=Korean2026.01 | 5.47 | 22 | |
| Phi2.8B tuned Koreanparameters=2.8B, tuning=Korean2026.01 | 5.83 | 18 | |
| Llama10.8B tuned Koreanparameters=10.8B, tuning=Korean2026.01 | 8.13 | 4 | |
| Llama2 13Bparameters=13B2026.01 | 8.53 | 15 | |
| Llama2 7Bparameters=7B2026.01 | 8.7 | 16 | |
| Phi2.8B tuned Koreanparameters=2.8B, tuning=Korean2026.01 | 8.72 | 3 | |
| FiD2026.01 | 9.51 | 22 | |
| FiD-RAG2026.01 | 9.66 | 19 | |
| RAG2026.01 | 9.74 | 26 | |
| FiD2026.01 | 10.49 | 10 | |
| Llama2 7Bparameters=7B2026.01 | 10.5 | 3 | |
| Llama2 13Bparameters=13B2026.01 | 10.51 | 3 | |
| FiD-RAG2026.01 | 10.52 | 4 | |
| RAG2026.01 | 10.55 | 7 |