KURE-Reranker-nano

KURE-Reranker-nano is a lightweight, approximately 150M-parameter Korean-English bilingual reranker. It jointly encodes a query and a candidate document and predicts a scalar relevance score.

Key Characteristics

• Strong Korean reranking: 87.57 average nDCG@10 across nine Korean benchmarks, the highest average among the evaluated models with fewer than 1B parameters and only 0.43 points below KURE-Reranker-base (1.7B).
• Long-document reranking: 79.66 nDCG@10 on MultiLongDocRetrieval (MLDR), the best among the evaluated models with fewer than 1B parameters.
• High throughput: 482.4 pairs per second averaged over nine benchmarks on an NVIDIA RTX A6000 48GB, about 19.8× Qwen3-Reranker-4B under the measurement conditions described below.
• Lightweight: Approximately 150M parameters, based on skt/A.X-Encoder-base.

Model Overview

Usage

Install the required libraries:

pip install "sentence-transformers>=5.1.0" "transformers>=4.55.4" torch

from sentence_transformers import CrossEncoder

model = CrossEncoder("nlpai-lab/KURE-Reranker-nano", max_length=8192)

query = "대한민국의 수도는 어디인가요?"
documents = [
    "대한민국의 수도는 서울입니다.",
    "The capital of South Korea is Seoul.",
    "바나나는 열대 지역에서 재배되는 과일입니다.",
]

results = model.rank(query, documents, batch_size=8, return_documents=True)
for result in results:
    print(f"{result['score']:.4f}\t{result['text']}")

pairs = [(query, document) for document in documents]
scores = model.predict(pairs, batch_size=8)

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_id = "nlpai-lab/KURE-Reranker-nano"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSequenceClassification.from_pretrained(model_id).to(device)
model.eval()

query = "대한민국의 수도는 어디인가요?"
documents = [
    "대한민국의 수도는 서울입니다.",
    "The capital of South Korea is Seoul.",
    "바나나는 열대 지역에서 재배되는 과일입니다.",
]
inputs = tokenizer(
    [query] * len(documents), documents,
    padding=True, truncation=True, max_length=8192, return_tensors="pt",
).to(device)

with torch.inference_mode():
    scores = model(**inputs).logits.squeeze(-1).float().cpu()

for index in torch.argsort(scores, descending=True).tolist():
    print(f"{scores[index]:.4f}\t{documents[index]}")

Korean Reranking Evaluation

Scores are nDCG@10, and throughput is pairs per second (PPS). Evaluation was conducted using the reranker-simple-benchmark implementation as a reference.

The figure plots mean nDCG@10 against mean PPS across the nine Korean benchmarks, including MLDR. Dot color marks model size; stars mark the KURE-Reranker models. PPS uses each model's own input limit and its best measured batch size, so it measures neither speed at equal token lengths nor end-to-end retrieval latency. The x-axis is a non-uniform log scale that stretches 20–26 and compresses 26–50 pairs/s so that nearby points stay readable.

Results — MTEB-ko-retrieval (9 subsets)

PPS is measured on a single NVIDIA RTX A6000 48GB per model and averaged over the nine benchmarks. For each benchmark, every model is timed on the same ~550 sampled query–document pairs (whole queries, fixed seed). Inputs are length-sorted and dynamically padded within each batch. Models run in bf16 with flash_attention_2, except KaLM-Embedding/KaLM-Reranker-V1-Large-R2, which uses sdpa because transformers does not support flash_attention_2 for T5Gemma2. Each query is reranked over its gold documents plus, as negatives, the BM25 top-50 minus the golds; BelebeleRetrieval uses its kor_Hang-kor_Hang subset.

jinaai/jina-reranker-v3 and jinaai/jina-reranker-v3.5 are listwise rerankers that cannot be compared under the same 8,192-token condition on MLDR, so they are excluded from the 9-subset table and their MLDR cells are left blank below.

Per-dataset NDCG@10

Per-dataset PPS

Batch size starts at 8 and doubles until an out-of-memory error. Samples are repeated to fill complete batches. After warmup, three full passes are timed with CUDA events. Throughput is the total number of processed pairs divided by the accumulated GPU forward time. The highest-throughput successful batch is reported. Tokenization, data loading, and CPU preprocessing are excluded. The batch-search approach is informed by the Ettin reranker speed benchmark.

Citation

@misc{kure-reranker-nano,
  title  = {KURE-Reranker-nano: A Lightweight Korean–English Bilingual Reranking Model},
  author = {Hong, Seongtae and Jang, Youngjoon and Son, Junyoung and Lee, Taemin and Lim, Heuiseok},
  year   = {2026},
  url    = {https://huggingface.co/nlpai-lab/KURE-Reranker-nano},
}