🔎 KURE-Sparse-base

KURE-Sparse-base is a Korean-English bilingual learned sparse retrieval model. It shows state-of-the-art performance among sparse models supporting Korean, averaging 0.7984 nDCG@10 on MTEB(kor, v2) Retrieval. It also outperforms 8B dense models, such as Qwen3-Embedding-8B and comsat-embed-ko-8b-preview. It is also served through the KURE API.

Key Characteristics

• Learned Sparse Retrieval (LSR): every vector lives in a fixed vocabulary, so documents can be indexed in an inverted index and matched terms are inspectable.
• Dual view: represented with 30,522 English pivot terms (projected from the multilingual encoder) + 250,002 source-language tokens.
• Long documents: up to 8,192 tokens.
• Sparse at both ends: on average 100 active dimensions per query and 1,897 per document.

Usage

Sentence Transformers

pip install -U sentence-transformers

from sentence_transformers import SparseEncoder

model = SparseEncoder("nlpai-lab/KURE-Sparse-base", trust_remote_code=True)

query = "훈민정음은 언제 만들어졌나요?"
documents = [
    "세종대왕은 1443년에 훈민정음을 창제하고 1446년에 이를 반포하였다.",
    "김치는 배추나 무를 소금에 절인 뒤 고춧가루와 젓갈을 넣어 발효시킨 음식이다.",
    "한라산은 해발 1,947m로 남한에서 가장 높은 산이며 제주도 중앙에 자리한다.",
]

query_embeddings = model.encode_query(query)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# torch.Size([280524]) torch.Size([3, 280524])

# Dot-product scores (higher is more relevant)
scores = model.similarity(query_embeddings, document_embeddings)
print(scores)
# tensor([[23.9621,  2.0522,  1.0599]])

API

KURE-Sparse-base is served at kure.kullm.ai. Sign in with GitHub there to get an API key.

import os

import requests

URL = "https://kure.kullm.ai/api/v1/sparse-embeddings"
HEADERS = {"Authorization": f"Bearer {os.environ['KURE_API_KEY']}"}

query = "훈민정음은 언제 만들어졌나요?"
documents = [
    "세종대왕은 1443년에 훈민정음을 창제하고 1446년에 이를 반포하였다.",
    "김치는 배추나 무를 소금에 절인 뒤 고춧가루와 젓갈을 넣어 발효시킨 음식이다.",
    "한라산은 해발 1,947m로 남한에서 가장 높은 산이며 제주도 중앙에 자리한다.",
]

# input_type adds the query or document markers. return_tokens=True also returns the term of each index
def encode(texts, input_type):
    response = requests.post(URL, headers=HEADERS, json={"model": "kure-sparse-base", "input": texts, "input_type": input_type})
    response.raise_for_status()
    return [dict(zip(e["indices"], e["values"])) for e in response.json()["data"]]

query_embedding = encode(query, "query")[0]
document_embeddings = encode(documents, "document")

# Dot-product scores (higher is more relevant)
scores = [sum(w * doc.get(i, 0.0) for i, w in query_embedding.items()) for doc in document_embeddings]
print([round(s, 4) for s in scores])
# [23.9825, 2.0616, 1.0396]

Evaluation

KURE-Sparse-base is evaluated on the nine MTEB(kor, v2) Retrieval tasks. We report nDCG@10.

Model	Type	Params	Avg	AutoRAG	PubHealthQA	Ko-StrategyQA	LawIRKo	SQuADKorV1	Belebele (ko-ko)	MrTidy	MLDR	MIRACL
KURE Models
nlpai-lab/KURE-Sparse-base	Sparse	0.6B	0.7984	0.9445	0.8099	0.8188	0.7244	0.9815	0.9703	0.6255	0.5716	0.7395
nlpai-lab/KURE-v3	Dense	1.1B	0.8027	0.9324	0.8639	0.8192	0.8158	0.9666	0.9673	0.6320	0.4781	0.7489
nlpai-lab/KURE-v2	Late-interaction	0.2B	0.8160	0.9718	0.8228	0.8070	0.7550	0.9846	0.9660	0.5974	0.7158	0.7237
nlpai-lab/KURE-Sparse-nano	Sparse	0.1B	0.7716	0.9360	0.7711	0.7924	0.7226	0.9774	0.9608	0.5618	0.5133	0.7091
nlpai-lab/KURE-v1	Dense	0.6B	0.7616	0.8708	0.8193	0.7999	0.7426	0.9357	0.9502	0.5909	0.4637	0.6816
nlpai-lab/KoE5	Dense	0.6B	0.7337	0.8434	0.8351	0.8001	0.7756	0.8980	0.9425	0.5841	0.3015	0.6235
Dense (Single-Vector) Models
nvidia/Nemotron-3-Embed-8B-BF16	Dense	8.0B	0.8127	0.9311	0.8898	0.8405	0.8481	0.9650	0.9714	0.6540	0.4785	0.7363
sionic-ai/comsat-embed-ko-8b-preview	Dense	7.6B	0.7927	0.8518	0.8871	0.8394	0.8164	0.9168	0.9853	0.6253	0.5157	0.6964
Qwen/Qwen3-Embedding-8B	Dense	7.6B	0.7826	0.8276	0.8721	0.8363	0.8171	0.9063	0.9824	0.6187	0.5046	0.6783
nvidia/Nemotron-3-Embed-1B-BF16	Dense	1.1B	0.7817	0.9074	0.8642	0.8136	0.7758	0.9505	0.9629	0.6084	0.4202	0.7325
Qwen/Qwen3-Embedding-4B	Dense	4.0B	0.7737	0.8431	0.8693	0.8270	0.7769	0.9044	0.9522	0.6076	0.5022	0.6803
perplexity-ai/pplx-embed-v1-4b	Dense	4.0B	0.7705	0.8321	0.9042	0.8326	0.7736	0.9402	0.9624	0.5778	0.4343	0.6775
microsoft/harrier-oss-v1-27b	Dense	27.0B	0.7667	0.8176	0.8971	0.8361	0.8737	0.9204	0.9546	0.5306	0.4046	0.6653
dragonkue/snowflake-arctic-embed-l-v2.0-ko	Dense	0.6B	0.7653	0.9093	0.8337	0.8050	0.7735	0.9447	0.9518	0.5712	0.4304	0.6685
codefuse-ai/F2LLM-v2-8B	Dense	7.6B	0.7638	0.7678	0.9380	0.8371	0.8405	0.8874	0.9513	0.6162	0.4047	0.6313
telepix/PIXIE-Rune-v1.5	Dense	0.6B	0.7618	0.8927	0.8426	0.8064	0.7705	0.9457	0.9617	0.5492	0.4482	0.6393
nvidia/llama-nemotron-embed-1b-v2	Dense	1.2B	0.7564	0.8708	0.8177	0.8055	0.7329	0.9345	0.9519	0.5988	0.3952	0.7004
dragonkue/BGE-m3-ko	Dense	0.6B	0.7547	0.8738	0.8155	0.7959	0.7322	0.9414	0.9503	0.6099	0.3899	0.6833
BAAI/bge-m3	Dense	0.6B	0.7509	0.8301	0.8041	0.7941	0.7174	0.9038	0.9315	0.6471	0.4287	0.7015
jinaai/jina-embeddings-v5-text-small-retrieval	Dense	0.6B	0.7419	0.8409	0.8664	0.8096	0.7483	0.8926	0.9346	0.5779	0.3315	0.6752
perplexity-ai/pplx-embed-v1-0.6b	Dense	0.6B	0.7390	0.8414	0.8255	0.8100	0.7444	0.9174	0.9393	0.5524	0.3179	0.7026
google/embeddinggemma-300m	Dense	0.3B	0.7254	0.8187	0.8170	0.8008	0.6850	0.9114	0.9430	0.5253	0.3844	0.6434
Qwen/Qwen3-Embedding-0.6B	Dense	0.6B	0.7140	0.8338	0.8107	0.7586	0.7526	0.8490	0.9124	0.4870	0.4199	0.6023
Late-interaction (Multi-Vector) Models
lightonai/mLateOn	Late-interaction	0.3B	0.7906	0.9392	0.8061	0.7905	0.6431	0.9803	0.9608	0.5817	0.7005	0.7135
perplexity-ai/pplx-embed-v1-late-0.6b	Late-interaction	0.6B	0.7381	0.8557	0.8089	0.7973	0.7285	0.9696	0.9548	0.5400	0.2816	0.7064
dragonkue/colbert-ko-0.1b	Late-interaction	0.1B	0.6776	0.9700	0.7482	0.7364	0.4475	0.9794	0.9644	0.3966	0.2872	0.5685
Sparse Models
omai-research/milco-650m	Sparse	0.6B	0.7824	0.9483	0.8219	0.8143	0.6724	0.9700	0.9623	0.6997	0.4318	0.7209
telepix/PIXIE-Splade-v1.5	Sparse	0.1B	0.7608	0.9412	0.7910	0.7993	0.7667	0.9687	0.9651	0.5132	0.4938	0.6085

Performance-Sparsity

Average active dimensions per vector and FLOPs, averaged over the nine tasks.

Model	Params	Query dims	Document dims	FLOPs
KURE-Sparse-base	0.6B	100	1,897	36.5
KURE-Sparse-nano	0.1B	147	843	23.0
omai-research/milco-650m	0.6B	105	1,167	32.3
telepix/PIXIE-Splade-v1.5	0.1B	117	1,541	37.3
yjoonjang/splade-ko-v1	0.1B	109	742	29.1
telepix/PIXIE-Splade-v1.0	0.1B	121	625	22.4
opensearch-project/opensearch-neural-sparse-encoding-multilingual-v1	0.2B	29	176	6.8

English Retrieval (BEIR, RTEB)

KURE-Sparse-base is also evaluated on two English benchmarks with nDCG@10: the 13 BEIR tasks (BEIR-13), and the eight public English tasks of RTEB outside the code domain. SparseUp runs at its own length limits (128 query and 512 document tokens), and the other models at up to 8,192 tokens. The RTEB average differs from the RTEB leaderboard score, which also covers private and code tasks.

BEIR-13

Model	Params	BEIR-13	FLOPs
KURE Models
KURE-Sparse-base	0.6B	0.5864	55.8
KURE-Sparse-nano	0.1B	0.5317	37.0
Sparse Models
utahnlp/laconic-1b	1.2B	0.5831	52.6
Linkup-Platform/linkup-sparseup-embed-v1	0.2B	0.5645	8.5
omai-research/milco-650m	0.6B	0.5545	20.6
opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte	0.1B	0.5459	1.7
naver/splade-v3	0.1B	0.5202	8.0
telepix/PIXIE-Splade-v1.5	0.1B	0.4149	125.6

Per-task nDCG@10

Model	Avg	ArguAna	ClimateFEVER	DBPedia	FEVER	FiQA	HotpotQA	NFCorpus	NQ	Quora	SCIDOCS	SciFact	TREC-COVID	Touche-2020
KURE Models
KURE-Sparse-base	0.5864	0.5896	0.3959	0.4837	0.9325	0.4720	0.8084	0.3678	0.6758	0.8252	0.1741	0.7183	0.8260	0.3535
KURE-Sparse-nano	0.5317	0.5075	0.2865	0.4214	0.9079	0.4080	0.7476	0.3340	0.5647	0.7825	0.1581	0.7120	0.7637	0.3183
Sparse Models
utahnlp/laconic-1b	0.5831	0.6127	0.3952	0.4705	0.8871	0.4427	0.8018	0.3958	0.6642	0.8072	0.2397	0.7596	0.8300	0.2737
Linkup-Platform/linkup-sparseup-embed-v1	0.5645	0.5648	0.3747	0.4150	0.8969	0.4687	0.7761	0.3523	0.5717	0.8656	0.1852	0.7376	0.7823	0.3478
omai-research/milco-650m	0.5545	0.6157	0.3079	0.4501	0.8353	0.4262	0.7780	0.3629	0.6495	0.8818	0.1704	0.7020	0.7476	0.2813
opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte	0.5459	0.5196	0.3123	0.4553	0.8598	0.4069	0.7158	0.3592	0.5817	0.8734	0.1675	0.7251	0.7330	0.3866
naver/splade-v3	0.5202	0.4880	0.2566	0.4514	0.8098	0.3810	0.6924	0.3630	0.5867	0.8138	0.1557	0.7171	0.7346	0.3119
telepix/PIXIE-Splade-v1.5	0.4149	0.2185	0.1302	0.3519	0.6475	0.2753	0.5756	0.3325	0.3858	0.6878	0.1468	0.6694	0.7068	0.2650

RTEB

Model	Params	Avg	FLOPs
KURE Models
nlpai-lab/KURE-Sparse-base	0.6B	0.5093	277.6
nlpai-lab/KURE-Sparse-nano	0.1B	0.4750	109.9
Sparse Models
utahnlp/laconic-1b	1.2B	0.5415	107.2
Linkup-Platform/linkup-sparseup-embed-v1	0.2B	0.4941	25.1
opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte	0.1B	0.4721	12.7
naver/splade-v3	0.1B	0.4596	35.0
omai-research/milco-650m	0.6B	0.4301	163.7
telepix/PIXIE-Splade-v1.5	0.1B	0.3630	391.3

Per-task nDCG@10

Model	Params	Avg	CUREv1	ChatDoctor	FinQA	AILACasedocs	LegalSummarization	HC3Finance	FinanceBench	AILAStatutes	FLOPs
KURE Models
nlpai-lab/KURE-Sparse-base	0.6B	0.5093	0.5996	0.3914	0.7838	0.2298	0.5494	0.5951	0.7525	0.1727	277.6
nlpai-lab/KURE-Sparse-nano	0.1B	0.4750	0.3591	0.4730	0.7042	0.2860	0.6399	0.4438	0.7005	0.1939	109.9
Sparse Models
utahnlp/laconic-1b	1.2B	0.5415	0.5340	0.5272	0.6727	0.3436	0.6064	0.5893	0.7449	0.3142	107.2
Linkup-Platform/linkup-sparseup-embed-v1	0.2B	0.4941	0.4010	0.5424	0.5805	0.2406	0.6430	0.5900	0.6852	0.2700	25.1
opensearch-project/opensearch-neural-sparse-encoding-doc-v3-gte	0.1B	0.4721	0.3004	0.5015	0.7344	0.2596	0.6554	0.4265	0.6855	0.2139	12.7
naver/splade-v3	0.1B	0.4596	0.3306	0.5036	0.6374	0.3153	0.5663	0.4118	0.6878	0.2241	35.0
omai-research/milco-650m	0.6B	0.4301	0.5674	0.2344	0.7399	0.2197	0.3078	0.5451	0.6513	0.1747	163.7
telepix/PIXIE-Splade-v1.5	0.1B	0.3630	0.3414	0.2178	0.5751	0.2082	0.5216	0.3042	0.5180	0.2181	391.3

Citation

@misc{kure-sparse,
  title  = {KURE-Sparse: a Korean-English bilingual learned sparse retriever},
  author = {Jang, Youngjoon and Son, Junyoung and Lee, Taemin and Hong, Seongtae and Lim, Heuiseok},
  year   = {2026},
  url    = {https://huggingface.co/nlpai-lab/KURE-Sparse-base},
}

@inproceedings{jang2025kure,
  title={KURE: Embedding Model for Korean-Specific Retrieval},
  author={Jang, Youngjoon and Son, Junyoung and Lee, Taemin and Hong, Seongtae and Park, JeongBae and Lim, Heuiseok},
  booktitle={Annual Conference on Human and Language Technology},
  pages={129--134},
  year={2025},
  organization={Human and Language Technology}
}

@inproceedings{nguyen2026milco,
  title={MILCO: Learned Sparse Retrieval Across Languages via a Multilingual Connector},
  author={Nguyen, Thong and Lei, Yibin and Ju, Jia-Huei and Yang, Eugene and Yates, Andrew},
  booktitle={International Conference on Learning Representations},
  year={2026}
}