Japanese CrossEncoder reranker models ranging from xsmall to large (plus BGE), evaluated on JQaRA, JaCWIR, MIRACL, and JSQuAD, with ready‑to‑use integration examples for sentence_transformers and HuggingFace.
Description in other languages
Statistics
Statistics as of Nov 16, 2026
Hugging Face downloads (last 30 days)
16,000
Hugging Face likes
7
Listed
May 13, 2024
Downloads (30-day window) over time
Show as a table
| Month | DL (30d) |
|---|---|
| Dec 2025 | 6,000 |
| Jan 2026 | 16,000 |
| Feb 2026 | 40,000 |
| Mar 2026 | 46,000 |
| Apr 2026 | 48,000 |
| May 2026 | 49,000 |
| Jun 2026 | 55,000 |
| Jul 2026 | 54,000 |
| Aug 2026 | 64,000 |
| Sep 2026 | 19,000 |
| Oct 2026 | 15,000 |
| Nov 2026 | 16,000 |
Downloads over the 30 days before the last snapshot of each month.
Popular in the same category
- japanese-reranker-cross-encoder-small-v1hotchpotch156kJapanese‑trained CrossEncoder rerankers ranging from xsmall (384) to large (1024) plus a BGE‑v2‑m3‑v1 model, with example code for fine‑tuning, inference, and benchmark scores on JQaRA, JaCWIR, MIRACL, and JSQuAD.
- ruri-v3-reranker-310mcl-nagoya137kRuri‑v3 Reranker is a robust Japanese text reranker built on ModernBERT‑Ja, supporting up to 8,192‑token sequences, a 100k‑token vocabulary, FlashAttention and a SentencePiece tokenizer, and it can be used via sentence‑transformers.
- japanese-reranker-xsmall-v2hotchpotch28kFast, lightweight Japanese Reranker v2 models (tiny, xsmall, small, base) with benchmark scores and GPU speeds, usable via sentence_transformers CrossEncoder and transformers ≥ v4.48 (optionally accelerated with flash‑attn) and also available in ONNX/quantized forms for CPU/ARM.
- ruri-reranker-smallcl-nagoya3kJapanese‑language reranker models built with Sentence Transformers (cross‑encoders) that can be loaded via trust_remote_code, benchmarked on JQaRA, JaCWIR, and MIRACL datasets, and available in small to large sizes from the hotchpotch organization.
- japanese-reranker-tiny-v2hotchpotch2kA compact, high‑speed Japanese reranker library (v2) offering tiny‑through‑base and cross‑encoder models with detailed latency and accuracy statistics, requiring Transformers ≥ 4.48 (optionally Flash‑Attention 2) and supporting ONNX/quantization for CPU/ARM deployments.
- japanese-reranker-small-v2hotchpotch2kJapanese‑reranker‑small‑v2 is a lightweight, fast Japanese reranker model series (v2) offering tiny to base variants that reach up to 0.89 average score with 2–15 s GPU inference, plus cross‑encoder options, and requires Hugging Face Transformers v4.48+ with optional Flash Attention 2 for acceleration.
Resources in similar research fields
- Llama-3.1-Swallow-8B-Instruct-v0.3tokyotech-llm5kLlama 3.1 Swallow is a Japanese‑enhanced series of 8B/70B Llama 3.1 models, trained via continual pre‑training and Japanese‑specific instruction fine‑tuning, with the latest 8B‑Instruct‑v0.3 setting state‑of‑the‑art results on Japanese MT‑Bench.Shared: Language Models
- Llama-3-ELYZA-JP-8Belyza4kLlama‑3‑ELYZA‑JP‑8B is a Japanese‑enhanced, 8‑billion‑parameter Llama 3 model by ELYZA, fine‑tuned for Japanese on Meta‑Llama‑3‑8B‑Instruct.Shared: Language Models
swallow-evaluation-instructswallow-llm34Swallow Project Post-learning Large-scale Language Model Evaluation FrameworkGitHubShared: Language Models
pretrained_doc2vec_jayagays25pretrained doc2vec models on Japanese WikipediaGitHubShared: Language Models
pl-bert-jakyamauchi102324A repository of Japanese Phoneme-Level BERTGitHubShared: Language Models
heron-vlm-leaderboardwandb6This project is a benchmarking tool for evaluating and comparing the performance of various Vision Language Models (VLMs). It uses two datasets: LLaVA-Bench-In-the-Wild and Japanese HERON Bench to measure model performance.GitHubShared: Language Models