metadata
language:
- ko
library_name: transformers
pipeline_tag: text-generation
Model Developers HyunseokLee, TaeyoungKim - (kaist alinlab, omnious.ai)
Input Models input text only.
Output Models generate text only.
Model Architecture
ko-ref-llama2-13b is an auto-regressive language model based on the LLaMA2 transformer architecture.
Base Model
Llama-2-13B
Training Dataset
Open dataset (Korean).
Training Objective
We trained the model to learn Korean corpus.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 41.32 |
| ARC (25-shot) | 48.38 |
| HellaSwag (10-shot) | 73.56 |
| MMLU (5-shot) | 34.83 |
| TruthfulQA (0-shot) | 35.82 |
| Winogrande (5-shot) | 69.14 |
| GSM8K (5-shot) | 0.0 |
| DROP (3-shot) | 27.53 |