cl-nagoya_ruri-v3-310m

---

language:

• ja tags:
• sentence-similarity
• feature-extraction base_model: cl-nagoya/ruri-v3-pt-310m widget: [] pipeline_tag: sentence-similarity license: apache-2.0 datasets:
• cl-nagoya/ruri-v3-dataset-ft

---

Ruri: Japanese General Text Embeddings

Ruri v3 is a general-purpose Japanese text embedding model built on top of ModernBERT-Ja. Ruri v3 offers several key technical advantages:

• State-of-the-art performance for Japanese text embedding tasks.
• Supports sequence lengths up to 8192 tokens
  • Previous versions of Ruri (v1, v2) were limited to 512.
• Expanded vocabulary of 100K tokens, compared to 32K in v1 and v2
  • The larger vocabulary make input sequences shorter, improving efficiency.
• Integrated FlashAttention, following ModernBERT's architecture
  • Enables faster inference and fine-tuning.
• Tokenizer based solely on SentencePiece
  • Unlike previous versions, which relied on Japanese-specific BERT tokenizers and required pre-tokenized input, Ruri v3 performs tokenization with SentencePiece only—no external word segmentation tool is required.

Model Series

We provide Ruri-v3 in several model sizes. Below is a summary of each model.

ID	#Param.	#Param. w/o Emb.	Dim.	#Layers	Avg. JMTEB
cl-nagoya/ruri-v3-30m	37M	10M	256	10	74.51
cl-nagoya/ruri-v3-70m	70M	31M	384	13	75.48
cl-nagoya/ruri-v3-130m	132M	80M	512	19	76.55
cl-nagoya/ruri-v3-310m	315M	236M	768	25	77.24

Usage

You can use our models directly with the transformers library v4.48.0 or higher:

pip install -U "transformers>=4.48.0" sentence-transformers

Additionally, if your GPUs support Flash Attention 2, we recommend using our models with Flash Attention 2.

pip install flash-attn --no-build-isolation

Then you can load this model and run inference.

import torch
import torch.nn.functional as F
from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
device = "cuda" if torch.cuda.is_available() else "cpu"
model = SentenceTransformer("cl-nagoya/ruri-v3-310m", device=device)

# Ruri v3 employs a 1+3 prefix scheme to distinguish between different types of text inputs:
# "" (empty string) is used for encoding semantic meaning.
# "トピック: " is used for classification, clustering, and encoding topical information.
# "検索クエリ: " is used for queries in retrieval tasks.
# "検索文書: " is used for documents to be retrieved.
sentences = [
    "川べりでサーフボードを持った人たちがいます",
    "サーファーたちが川べりに立っています",
    "トピック: 瑠璃色のサーファー",
    "検索クエリ: 瑠璃色はどんな色？",
    "検索文書: 瑠璃色（るりいろ）は、紫みを帯びた濃い青。名は、半貴石の瑠璃（ラピスラズリ、英: lapis lazuli）による。JIS慣用色名では「こい紫みの青」（略号 dp-pB）と定義している[1][2]。",
]

embeddings = model.encode(sentences, convert_to_tensor=True)
print(embeddings.size())
# [5, 768]

similarities = F.cosine_similarity(embeddings.unsqueeze(0), embeddings.unsqueeze(1), dim=2)
print(similarities)
# [[1.0000, 0.9603, 0.8157, 0.7074, 0.6916],
#  [0.9603, 1.0000, 0.8192, 0.7014, 0.6819],
#  [0.8157, 0.8192, 1.0000, 0.8701, 0.8470],
#  [0.7074, 0.7014, 0.8701, 1.0000, 0.9746],
#  [0.6916, 0.6819, 0.8470, 0.9746, 1.0000]]

Benchmarks

JMTEB

Evaluated with JMTEB.

Model	#Param.	Avg.	Retrieval	STS	Classfification	Reranking	Clustering	PairClassification
Ruri-v3-30m	37M	74.51	78.08	82.48	74.80	93.00	52.12	62.40
Ruri-v3-70m	70M	75.48	79.96	79.82	76.97	93.27	52.70	61.75
Ruri-v3-130m	132M	76.55	81.89	79.25	77.16	93.31	55.36	62.26
Ruri-v3-310m (this model)	315M	77.24	81.89	81.22	78.66	93.43	55.69	62.60
sbintuitions/sarashina-embedding-v1-1b	1.22B	75.50	77.61	82.71	78.37	93.74	53.86	62.00
PLaMo-Embedding-1B	1.05B	76.10	79.94	83.14	77.20	93.57	53.47	62.37
OpenAI/text-embedding-ada-002	-	69.48	64.38	79.02	69.75	93.04	48.30	62.40
OpenAI/text-embedding-3-small	-	70.86	66.39	79.46	73.06	92.92	51.06	62.27
OpenAI/text-embedding-3-large	-	73.97	74.48	82.52	77.58	93.58	53.32	62.35
pkshatech/GLuCoSE-base-ja	133M	70.44	59.02	78.71	76.82	91.90	49.78	66.39
pkshatech/GLuCoSE-base-ja-v2	133M	72.23	73.36	82.96	74.21	93.01	48.65	62.37
retrieva-jp/amber-base	130M	72.12	73.40	77.81	76.14	93.27	48.05	64.03
retrieva-jp/amber-large	315M	73.22	75.40	79.32	77.14	93.54	48.73	60.97
sentence-transformers/LaBSE	472M	64.70	40.12	76.56	72.66	91.63	44.88	62.33
intfloat/multilingual-e5-small	118M	69.52	67.27	80.07	67.62	93.03	46.91	62.19
intfloat/multilingual-e5-base	278M	70.12	68.21	79.84	69.30	92.85	48.26	62.26
intfloat/multilingual-e5-large	560M	71.65	70.98	79.70	72.89	92.96	51.24	62.15
Ruri-Small	68M	71.53	69.41	82.79	76.22	93.00	51.19	62.11
Ruri-Small v2	68M	73.30	73.94	82.91	76.17	93.20	51.58	62.32
Ruri-Base	111M	71.91	69.82	82.87	75.58	92.91	54.16	62.38
Ruri-Base v2	111M	72.48	72.33	83.03	75.34	93.17	51.38	62.35
Ruri-Large	337M	73.31	73.02	83.13	77.43	92.99	51.82	62.29
Ruri-Large v2	337M	74.55	76.34	83.17	77.18	93.21	52.14	62.27

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: cl-nagoya/ruri-v3-pt-310m
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768
• Similarity Function: Cosine Similarity
• Language: Japanese
• License: Apache 2.0
• Paper: https://arxiv.org/abs/2409.07737

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Citation

@misc{
  Ruri,
  title={{Ruri: Japanese General Text Embeddings}}, 
  author={Hayato Tsukagoshi and Ryohei Sasano},
  year={2024},
  eprint={2409.07737},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2409.07737}, 
}

License

This model is published under the Apache License, Version 2.0.