---
language:
- ko
pipeline_tag: text-generation
tags:
- Yi
- dpo
---

# Yi-6b-dpo

### Model Details
- Base Model: [beomi/Yi-Ko-6B](https://huggingface.co/beomi/Yi-Ko-6B)

### Datasets
- sampling and translate [Open-Orca/SlimOrca](https://huggingface.co/datasets/Open-Orca/SlimOrca)
- sampling and translate [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf)

### Benchmark
- SOTA model under 7B as of Dec 20, 2023 (https://huggingface.co/spaces/upstage/open-ko-llm-leaderboard).

| Model | Average |Ko-ARC | Ko-HellaSwag | Ko-MMLU | Ko-TruthfulQA | Ko-CommonGen V2 |
| --- | --- | --- | --- | --- | --- | --- |
| **hyeogi/Yi-6b-dpo-v0.2 (Ours)** | **52.63** | 41.72 | 52.96 | 46.69 | 52.38 | 69.42 |
| [hyeogi/Yi-6b-dpo-v0.1(Ours)](https://huggingface.co/hyeogi/Yi-6b-dpo-v0.1) | 51.38 | 41.3 | 52.23 | 45.34 | 54.03 | 63.99 |
| [Minirecord/Mini_DPO_7b_01](https://huggingface.co/Minirecord/Mini_DPO_7b_01) | 50.47 | 48.29 | 54.68 | 46.7 | 47.78 | 54.9

![image/png](https://cdn-uploads.huggingface.co/production/uploads/656e98a02c331f3e079e427f/wJ2es4j8Xemfv2yafIFp9.png)