Generative Modeling on Nagamese Conversational Corpus (test)
3.85Perplexity (PPL)NagaLLaMA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NagaLLaMAtraining_mode=fine-tuned2025.12 | 3.85 | 20.77 | |
| Llama 3.2-3Bevaluation_protocol=few-shot, parameters=3B2025.12 | 96.76 | 11.28 |
Loading the SOTA2 catalog…