LLM-jp-3 440M Instruct3
LLM-jp-3 440M Instruct3 is the 440M preference-aligned instruction release in LLM-jp's generation-3 size series. The underlying model was exposed to 2.1 trillion pretraining tokens.
Research release
Preference data targets helpfulness and safety
Instruct3 applies DPO after supervised fine-tuning. The card names aya-ja-evol-inst for helpfulness and ac-self-inst for safety preferences. The update changes alignment, while retaining the same model size and context length.LLM-jp-3 440M Instruct3 official model card
Review response behavior as well as accuracy
Practical guidance. For the 440M model, check whether preference alignment changes both useful answers and refusals. Larger models in the series cannot establish the behavior of this checkpoint.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | llm-jp-3-440m-instruct3 |
| Context window | 4,096 tokens Maximum sequence length stated in the model card. LLM-jp-3 440M Instruct3 official model card |
| Architecture | 16 layers; hidden size 1,024; 8 attention heads.LLM-jp-3 440M Instruct3 official model card |
| Parameters | 447,052,800 total, from the card's embedding and non-embedding counts.LLM-jp-3 440M Instruct3 official model card |
| License | Apache 2.0; research weights.LLM-jp-3 440M Instruct3 official model card |
Related models
- LLM-jp-3 440M Base
LLM-jp / NII LLMC · LLM-jp-3
LLM-jp-3 440M Base is a pretrained text model with 4,096-token context and Apache 2.0 weights.
Launch date not confirmedResearch releaseDownloadable - LLM-jp-3 440M Instruct2
LLM-jp / NII LLMC · LLM-jp-3
LLM-jp-3 440M Instruct2 is a supervised instruction release with 4,096-token context and Apache 2.0 weights.
Launch date not confirmedResearch releaseDownloadable