Dialogue Response Generation on KEEM memories 1.0 (test)
4.56PerplexityLlama10.8B tuned Korean
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Llama10.8B tuned Koreanparameters=10.8B, tuning=Korean2026.01 | 4.56 | 74 | |
| Phi2.8B tuned Koreanparameters=2.8B, tuning=Korean2026.01 | 4.61 | 79 | |
| Llama2 13Bparameters=13B2026.01 | 6.89 | 82 | |
| Llama2 7Bparameters=7B2026.01 | 6.99 | 81 | |
| FiD2026.01 | 7.88 | 68 | |
| FiD-RAG2026.01 | 7.9 | 77 | |
| RAG2026.01 | 8.1 | 67 |