llm-jp-3-440m-instruct3: benchmark results
LLM-jp 3 instruction-tuned release developed by NII's LLM-jp project, using Llama architecture and a 4,096-token context window. The named instruction-tuning generation remains distinct from the pretrained checkpoint and other instruct generations. Provider: National Institute of Informatics. Access: Open.
Aggregate Rating 1267 ± 42, rank #9097 of 10916 rated models, from 62 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Japanese LLM - Xlsum JA Rouge1 | 27.43 | Score (%) | 64.3 |
| Open Japanese LLM - Xlsum JA RougeLsum | 22.35 | Score (%) | 53.6 |
| Open Japanese LLM - Xlsum JA Bert Score JA F1 | 67.8 | Score (%) | 33.9 |
| Open Japanese LLM - Wiki Coreference SET F1 | 1.35 | Score (%) | 30.5 |
| Open Japanese LLM - SUM | 7.86 | Score (%) | 29.7 |
| Open Japanese LLM - Xlsum JA Rouge2 | 7.86 | Score (%) | 29.7 |
| Open Japanese LLM - Wiki NER SET F1 | 1.77 | Score (%) | 27 |
| pfgen-bench - Completion Mode - Truthfulness | 0.61 | Truthfulness Score | 26.7 |
| pfgen-bench - Completion Mode - Score | 0.36 | pfgen Score (mean of three) | 24.3 |
| pfgen-bench - Completion Mode - Helpfulness | 0.04 | Helpfulness Score | 23.3 |
| pfgen-bench - Completion Mode - Fluency | 0.44 | Fluency Score | 23 |
| Open Japanese LLM - Xlsum JA Bleu JA | 1.76 | Score (%) | 22.1 |
Interactive version: theaggregate.ai/model?slug=llm-jp-3-440m-instruct3 · How It Works · Data refreshed daily, snapshot 2026-11-08.