A Japanese-specialized language model developed by CyberAgent. Optimized for Japanese processing.
Parameters
Undisclosed
Context Window
License
Proprietary
Release Date
2026-03-19
API Pricing
API pricing for this model is not yet available
Strengths
- •Method robust against multimodal attacks
- •Improves model robustness via adversarial training
- •Enhances security in vision-language models
Weaknesses
- •Training is computationally expensive
- •Requires complex data augmentation pipelines
- •May slightly reduce clean data performance
Use Cases
- •Securing image-text models against attacks
- •Research on model robustness
- •Developing resilient AI systems