JaColBERTv2 is a Japanese‑only ColBERT‑based retrieval model trained with knowledge distillation on MMarco (31 negatives per positive, 250k steps, batch 32) that currently outperforms multilingual‑e5‑large, BGE‑M3, and JaColBERT, with a full evaluation pending.
Description in other languages
Statistics
Statistics as of Nov 16, 2026
Hugging Face downloads (last 30 days)
2,000
Hugging Face likes
17
Listed
Dec 15, 2025
Downloads (30-day window) over time
Show as a table
| Month | DL (30d) |
|---|---|
| Dec 2025 | 15,000 |
| Jan 2026 | 10,000 |
| Feb 2026 | 26,000 |
| Mar 2026 | 23,000 |
| Apr 2026 | 13,000 |
| May 2026 | 11,000 |
| Jun 2026 | 7,000 |
| Jul 2026 | 7,000 |
| Aug 2026 | 7,000 |
| Sep 2026 | 7,000 |
| Oct 2026 | 6,000 |
| Nov 2026 | 2,000 |
Downloads over the 30 days before the last snapshot of each month.
Popular in the same category
- ruri-v3-310mcl-nagoya634kRuri v3 is a state‑of‑the‑art Japanese text‑embedding model built on ModernBERT‑Ja, supporting up to 8,192‑token inputs, a 100K‑token vocabulary, FlashAttention‑accelerated inference, and multiple size variants for fast sentence‑transformer usage.
- plamo-embedding-1bpfnet486kPLaMo‑Embedding‑1B is a Japanese text‑embedding model from Preferred Networks that converts Japanese text into vectors for information retrieval, classification, and clustering, shows strong performance on the JMTEB benchmark, and is freely available under an Apache v2.0 license.
- ruri-v3-30mcl-nagoya126kRuri v3 is a state‑of‑the‑art Japanese text embedding model built on ModernBERT‑Ja, supporting up to 8,192 tokens, a 100 k‑token vocabulary, FlashAttention acceleration, and multiple sizes from 37 M to 315 M parameters.
- ruri-v3-130mcl-nagoya93kRuri v3 is a state‑of‑the‑art Japanese text‑embedding model built on ModernBERT‑Ja, supporting up to 8192‑token sequences, a 100K‑token vocabulary, FlashAttention, and released in sizes from 30 M to 310 M parameters for use with sentence‑transformers.
- GLuCoSE-base-japkshatech31kGLuCoSE is a Japanese sentence‑embedding model built on LUKE that outputs 768‑dim mean‑pooled vectors (up to 512 tokens) trained on web and NLI/search data, achieving 0.864 Spearman and 0.818 Pearson on similarity benchmarks.
- ruri-v3-70mcl-nagoya14kRuri v3 delivers high‑performance Japanese text embeddings up to 8192 tokens, a 100k‑token vocabulary, FlashAttention support, and multiple model sizes (30 m–310 m) for efficient inference and fine‑tuning via sentence‑transformers.
Resources in similar research fields
resemblatuem74Resembla: Word-based Japanese similar sentence search libraryGitHubShared: Semantic Similarity・Semantic Text Processing- ruri-base-v2cl-nagoya2kThe repository delivers new Japanese text‑embedding v3 models (30M‑310M params, 8 k‑token max, JMTEB 74‑77) that can be loaded with sentence_transformers (using query/passage prefixes) and includes benchmark comparisons with supervised/unsupervised SimCSE, LaBSE, E5, and other models.Shared: Semantic Similarity・Semantic Text Processing
JapaneseWordSimilarityDatasettmu-nlp103Japanese Word Similarity DatasetGitHubShared: Semantic Similarity・Semantic Text Processing- japanese-splade-v2hotchpotch4kHigh‑performance Japanese SPLADE v2 enables sparse‑vector conversion and inference through a WebUI demo, trains with YAST, offers YASEM embedding, and reports JMTEB benchmark results.Shared: Semantic Text Processing
wikipedia-passages-jawiki-embeddings-utilshotchpotch12wikipedia 日本語の文を、各種日本語の embeddings や faiss index へと変換するスクリプト等。GitHubShared: Semantic Text Processing- ruri-largecl-nagoya8kA collection of release‑ready Ruri v3 Japanese text embedding models (30m–310m), complete with SentenceTransformer usage tips, query/passage prefixes, and JMTEB benchmark results showing how they compare to other Japanese and multilingual embeddings.Shared: Semantic Text Processing