AI ExplorerAI Explorer
ToolsCategoriesSitesLLMsCompareAI QuizAlternativesPremium

—

AI Tools

—

Sites & Blogs

—

LLMs & Models

—

Categories

AI Explorer

Find and compare the best artificial intelligence tools for your projects.

Made within France

Explore

  • All tools
  • Sites & Blogs
  • LLMs & Models
  • Compare
  • Chatbots
  • AI Images
  • Code & Dev

Company

  • Premium
  • About
  • Contact
  • Blog

Legal

  • Legal notice
  • Privacy
  • Terms

© 2026 AI Explorer. All rights reserved.

HomeLLMsKURE v1

KURE v1

by nlpai-lab

Open source · 45k downloads · 84 likes

2.4
(84 reviews)EmbeddingAPI & Local
About

KURE v1 is an embedding model developed by the NLP&AI Lab at Korea University, specializing in Korean text retrieval. It outperforms most existing multilingual models for this task and stands out as one of the best publicly available Korean retrieval models. Designed to work with both Korean and English queries and documents, it excels particularly in scenarios requiring nuanced language understanding. Its use cases include document retrieval, question-answering systems, and semantic analysis of long texts. What sets it apart is its training on specific Korean data and its architecture optimized to maximize retrieval result accuracy.

Documentation

🔎 KURE-v1

Introducing Korea University Retrieval Embedding model, KURE-v1 It has shown remarkable performance in Korean text retrieval, speficially overwhelming most multilingual embedding models.
To our knowledge, It is one of the best publicly opened Korean retrieval models.

For details, visit the KURE repository


Model Versions

Model NameDimensionSequence LengthIntroduction
KURE-v110248192Fine-tuned BAAI/bge-m3 with Korean data via CachedGISTEmbedLoss
KoE51024512Fine-tuned intfloat/multilingual-e5-large with ko-triplet-v1.0 via CachedMultipleNegativesRankingLoss

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub.

  • Developed by: NLP&AI Lab
  • Language(s) (NLP): Korean, English
  • License: MIT
  • Finetuned from model: BAAI/bge-m3

Example code

Install Dependencies

First install the Sentence Transformers library:

Bash
pip install -U sentence-transformers

Python code

Then you can load this model and run inference.

Python
from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("nlpai-lab/KURE-v1")

# Run inference
sentences = [
    '헌법과 법원조직법은 어떤 방식을 통해 기본권 보장 등의 다양한 법적 모색을 가능하게 했어',
    '4. 시사점과 개선방향 앞서 살펴본 바와 같이 우리 헌법과 「법원조직 법」은 대법원 구성을 다양화하여 기본권 보장과 민주주의 확립에 있어 다각적인 법적 모색을 가능하게 하는 것을 근본 규범으로 하고 있다. 더욱이 합의체로서의 대법원 원리를 채택하고 있는 것 역시 그 구성의 다양성을 요청하는 것으로 해석된다. 이와 같은 관점에서 볼 때 현직 법원장급 고위법관을 중심으로 대법원을 구성하는 관행은 개선할 필요가 있는 것으로 보인다.',
    '연방헌법재판소는 2001년 1월 24일 5:3의 다수견해로 「법원조직법」 제169조 제2문이 헌법에 합치된다는 판결을 내렸음 ○ 5인의 다수 재판관은 소송관계인의 인격권 보호, 공정한 절차의 보장과 방해받지 않는 법과 진실 발견 등을 근거로 하여 텔레비전 촬영에 대한 절대적인 금지를 헌법에 합치하는 것으로 보았음 ○ 그러나 나머지 3인의 재판관은 행정법원의 소송절차는 특별한 인격권 보호의 이익도 없으며, 텔레비전 공개주의로 인해 법과 진실 발견의 과정이 언제나 위태롭게 되는 것은 아니라면서 반대의견을 제시함 ○ 왜냐하면 행정법원의 소송절차에서는 소송당사자가 개인적으로 직접 심리에 참석하기보다는 변호사가 참석하는 경우가 많으며, 심리대상도 사실문제가 아닌 법률문제가 대부분이기 때문이라는 것임 □ 한편, 연방헌법재판소는 「연방헌법재판소법」(Bundesverfassungsgerichtsgesetz: BVerfGG) 제17a조에 따라 제한적이나마 재판에 대한 방송을 허용하고 있음 ○ 「연방헌법재판소법」 제17조에서 「법원조직법」 제14절 내지 제16절의 규정을 준용하도록 하고 있지만, 녹음이나 촬영을 통한 재판공개와 관련하여서는 「법원조직법」과 다른 내용을 규정하고 있음',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# Results for KURE-v1
# tensor([[1.0000, 0.6967, 0.5306],
#         [0.6967, 1.0000, 0.4427],
#         [0.5306, 0.4427, 1.0000]])

Training Details

Training Data

KURE-v1

  • Korean query-document-hard_negative(5) data
  • 2,000,000 examples

Training Procedure

  • loss: Used CachedGISTEmbedLoss by sentence-transformers
  • batch size: 4096
  • learning rate: 2e-05
  • epochs: 1

Evaluation

Metrics

  • Recall, Precision, NDCG, F1

Benchmark Datasets

  • Ko-StrategyQA: 한국어 ODQA multi-hop 검색 데이터셋 (StrategyQA 번역)
  • AutoRAGRetrieval: 금융, 공공, 의료, 법률, 커머스 5개 분야에 대해, pdf를 파싱하여 구성한 한국어 문서 검색 데이터셋
  • MIRACLRetrieval: Wikipedia 기반의 한국어 문서 검색 데이터셋
  • PublicHealthQA: 의료 및 공중보건 도메인에 대한 한국어 문서 검색 데이터셋
  • BelebeleRetrieval: FLORES-200 기반의 한국어 문서 검색 데이터셋
  • MrTidyRetrieval: Wikipedia 기반의 한국어 문서 검색 데이터셋
  • MultiLongDocRetrieval: 다양한 도메인의 한국어 장문 검색 데이터셋
  • XPQARetrieval: 다양한 도메인의 한국어 문서 검색 데이터셋

Results

아래는 모든 모델의, 모든 벤치마크 데이터셋에 대한 평균 결과입니다. 자세한 결과는 KURE Github에서 확인하실 수 있습니다.

Top-k 1

ModelAverage Recall_top1Average Precision_top1Average NDCG_top1Average F1_top1
nlpai-lab/KURE-v10.526400.605510.605510.55784
dragonkue/BGE-m3-ko0.523610.603940.603940.55535
BAAI/bge-m30.517780.598460.598460.54998
Snowflake/snowflake-arctic-embed-l-v2.00.512460.593840.593840.54489
nlpai-lab/KoE50.501570.577900.577900.53178
intfloat/multilingual-e5-large0.500520.577270.577270.53122
jinaai/jina-embeddings-v30.482870.560680.560680.51361
BAAI/bge-multilingual-gemma20.479040.554720.554720.50916
intfloat/multilingual-e5-large-instruct0.478420.554350.554350.50826
intfloat/multilingual-e5-base0.469500.544900.544900.49947
intfloat/e5-mistral-7b-instruct0.467720.543940.543940.49781
Alibaba-NLP/gte-multilingual-base0.464690.537440.537440.49353
Alibaba-NLP/gte-Qwen2-7B-instruct0.466330.536250.536250.49429
openai/text-embedding-3-large0.448840.516880.516880.47572
Salesforce/SFR-Embedding-2_R0.437480.508150.508150.46504
upskyy/bge-m3-korean0.431250.502450.502450.45945
jhgan/ko-sroberta-multitask0.337880.384970.384970.35678

Top-k 3

ModelAverage Recall_top1Average Precision_top1Average NDCG_top1Average F1_top1
nlpai-lab/KURE-v10.686780.287110.655380.39835
dragonkue/BGE-m3-ko0.678340.283850.649500.39378
BAAI/bge-m30.675260.283740.645560.39291
Snowflake/snowflake-arctic-embed-l-v2.00.671280.281930.640420.39072
intfloat/multilingual-e5-large0.658070.277770.628220.38423
nlpai-lab/KoE50.651740.273290.623690.37882
BAAI/bge-multilingual-gemma20.644150.274160.611050.37782
jinaai/jina-embeddings-v30.641160.271650.609540.37511
intfloat/multilingual-e5-large-instruct0.643530.270400.607900.37453
Alibaba-NLP/gte-multilingual-base0.637440.264040.596950.36764
Alibaba-NLP/gte-Qwen2-7B-instruct0.631630.259370.592370.36263
intfloat/multilingual-e5-base0.620990.261440.591790.36203
intfloat/e5-mistral-7b-instruct0.620870.261440.589170.36188
openai/text-embedding-3-large0.610350.253560.573290.35270
Salesforce/SFR-Embedding-2_R0.600010.252530.563460.34952
upskyy/bge-m3-korean0.592150.250760.557220.34623
jhgan/ko-sroberta-multitask0.469300.189940.432930.26696

Top-k 5

ModelAverage Recall_top1Average Precision_top1Average NDCG_top1Average F1_top1
nlpai-lab/KURE-v10.738510.191300.674790.29903
dragonkue/BGE-m3-ko0.725170.187990.666920.29401
BAAI/bge-m30.729540.189750.666150.29632
Snowflake/snowflake-arctic-embed-l-v2.00.729620.188750.662360.29542
nlpai-lab/KoE50.708200.182870.644990.28628
intfloat/multilingual-e5-large0.701240.183160.644020.28588
BAAI/bge-multilingual-gemma20.702580.185560.633380.28851
jinaai/jina-embeddings-v30.699330.182560.631330.28505
intfloat/multilingual-e5-large-instruct0.690180.178380.624860.27933
Alibaba-NLP/gte-multilingual-base0.693650.177890.618960.27879
intfloat/multilingual-e5-base0.672500.174060.611190.27247
Alibaba-NLP/gte-Qwen2-7B-instruct0.674470.171140.609520.26943
intfloat/e5-mistral-7b-instruct0.674490.174840.609350.27349
openai/text-embedding-3-large0.663650.170040.593890.26677
Salesforce/SFR-Embedding-2_R0.656220.170180.584940.26612
upskyy/bge-m3-korean0.654770.170150.580730.26589
jhgan/ko-sroberta-multitask0.531360.132640.458790.20976

Top-k 10

ModelAverage Recall_top1Average Precision_top1Average NDCG_top1Average F1_top1
nlpai-lab/KURE-v10.796820.106240.694730.18524
dragonkue/BGE-m3-ko0.784500.104920.687480.18288
BAAI/bge-m30.791950.105920.687230.18456
Snowflake/snowflake-arctic-embed-l-v2.00.786690.104620.681890.18260
intfloat/multilingual-e5-large0.759020.101470.663700.17693
nlpai-lab/KoE50.752960.099370.660120.17369
BAAI/bge-multilingual-gemma20.761530.103640.653300.18003
jinaai/jina-embeddings-v30.762770.102400.652900.17843
intfloat/multilingual-e5-large-instruct0.748510.098880.644510.17283
Alibaba-NLP/gte-multilingual-base0.756310.099380.640250.17363
Alibaba-NLP/gte-Qwen2-7B-instruct0.740920.096070.632580.16847
intfloat/multilingual-e5-base0.735120.097170.632160.16977
intfloat/e5-mistral-7b-instruct0.737950.097770.630760.17078
openai/text-embedding-3-large0.729460.095710.616700.16739
Salesforce/SFR-Embedding-2_R0.716620.095460.605890.16651
upskyy/bge-m3-korean0.718950.095830.602580.16712
jhgan/ko-sroberta-multitask0.612250.078260.486870.13757

Citation

If you find our paper or models helpful, please consider cite as follows:

Text
@misc{KURE,
  publisher = {Youngjoon Jang, Junyoung Son, Taemin Lee},
  year = {2024},
  url = {https://github.com/nlpai-lab/KURE}
},

@misc{KoE5,
  author = {NLP & AI Lab and Human-Inspired AI research},
  title = {KoE5: A New Dataset and Model for Improving Korean Embedding Performance},
  year = {2024},
  publisher = {Youngjoon Jang, Junyoung Son, Taemin Lee},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/nlpai-lab/KoE5}},
}
Capabilities & Tags
sentence-transformerssafetensorsxlm-robertasentence-similarityfeature-extractiongenerated_from_trainertext-embeddings-inferenceendpoints_compatible
Links & Resources
Specifications
CategoryEmbedding
AccessAPI & Local
LicenseOpen Source
PricingOpen Source
Rating
2.4

Try KURE v1

Access the model directly