---
license: mit
language:
- ko
- en
pipeline_tag: text-classification
---
# Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->

This modelcard aims to be a base template for new models. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/modelcard_template.md?plain=1).
### **한국어 Reranker** 개발을 위한 파인튜닝 가이드를 제시합니다.
ko-reranker는 [BAAI/bge-reranker-larger](https://huggingface.co/BAAI/bge-reranker-large) 기반 한국어 데이터에 대한 fine-tuned model 입니다. 

- - -

## 0. Usage
- #### <span style="#FF69B4;"> Reranker는 임베딩 모델과 달리 질문과 문서를 입력으로 사용하며 임베딩 대신 유사도를 직접 출력합니다.</span>
- #### <span style="#FF69B4;"> Reranker에 질문과 구절을 입력하면 연관성 점수를 얻을 수 있습니다.</span>
- #### <span style="#FF69B4;"> Reranker는 CrossEntropy loss를 기반으로 최적화되므로 관련성 점수가 특정 범위에 국한되지 않습니다.</span>

## 1. Backgound
- #### <span style="#FF69B4;"> **컨택스트 순서가 정확도에 영향 준다**([Lost in Middel, *Liu et al., 2023*](https://arxiv.org/pdf/2307.03172.pdf)) </span>

- #### <span style="#FF69B4;"> [Reranker 사용해야 하는 이유](https://www.pinecone.io/learn/series/rag/rerankers/)</span>
    - 현재 LLM은 context 많이 넣는다고 좋은거 아님, relevant한게 상위에 있어야 정답을 잘 말해준다
    - Semantic search에서 사용하는 similarity(relevant) score가 정교하지 않다. (즉, 상위 랭커면 하위 랭커보다 항상 더 질문에 유사한 정보가 맞아?) 
        * Embedding은 meaning behind document를 가지는 것에 특화되어 있다. 
        * 질문과 정답이 의미상 같은건 아니다. ([Hypothetical Document Embeddings](https://medium.com/prompt-engineering/hyde-revolutionising-search-with-hypothetical-document-embeddings-3474df795af8))
        * ANNs([Approximate Nearest Neighbors](https://towardsdatascience.com/comprehensive-guide-to-approximate-nearest-neighbors-algorithms-8b94f057d6b6)) 사용에 따른 패널티

- - -

## 2. Reranker models

- #### <span style="#FF69B4;"> [Cohere] [Reranker](https://txt.cohere.com/rerank/)</span>
- #### <span style="#FF69B4;"> [BAAI] [bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large)</span>
- #### <span style="#FF69B4;"> [BAAI] [bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base)</span>

- - -

## 3. Dataset

- #### <span style="#FF69B4;"> [msmarco-triplets](https://github.com/microsoft/MSMARCO-Passage-Ranking) </span>
    - (Question, Answer, Negative)-Triplets from MS MARCO Passages dataset, 499,184 samples
    - 해당 데이터 셋은 영문으로 구성되어 있습니다.
    - Amazon Translate 기반으로 번역하여 활용하였습니다.
    
- - -

## 4. Performance
| Model                     | has-right-in-contexts | mrr (mean reciprocal rank) |
|:---------------------------|:-----------------:|:--------------------------:|
| without-reranker (default)| 0.93 | 0.80 |
| with-reranker (bge-reranker-large)| 0.95 | 0.84 |
| **with-reranker (fine-tuned using korean)** | **0.96** | **0.87** |

- **evaluation set**:
```code
./dataset/evaluation/eval_dataset.csv
```
- **training parameters**: 

```json
{
    "learning_rate": 5e-6,
    "fp16": True,
    "num_train_epochs": 3,
    "per_device_train_batch_size": 1,
    "gradient_accumulation_steps": 32,
    "train_group_size": 3,
    "max_len": 512,
    "weight_decay": 0.01,
}
```

- - -

## 5. Acknowledgement
- <span style="#FF69B4;"> Part of the code is developed based on [FlagEmbedding](https://github.com/FlagOpen/FlagEmbedding/tree/master?tab=readme-ov-file) and [KoSimCSE-SageMaker](https://github.com/daekeun-ml/KoSimCSE-SageMaker/tree/7de6eefef8f1a646c664d0888319d17480a3ebe5).</span>

- - -

## 6. Citation
- <span style="#FF69B4;"> If you find this repository useful, please consider giving a star ⭐ and citation</span>

- - -

## 7. Contributors:
- <span style="#FF69B4;"> **Dongjin Jang, Ph.D.** (AWS AI/ML Specislist Solutions Architect) | [Mail](mailto:dongjinj@amazon.com) | [Linkedin](https://www.linkedin.com/in/dongjin-jang-kr/) | [Git](https://github.com/dongjin-ml) | </span>

- - -

## 8. License
- <span style="#FF69B4;"> FlagEmbedding is licensed under the [MIT License](https://github.com/aws-samples/aws-ai-ml-workshop-kr/blob/master/LICENSE). </span>