Ricoh develops Japanese LLM with guardrail function that judges the safety of generative AI input and output
Ricoh announced the development of an improved Japanese LLM that integrates a guardrail function to detect harmful information in generative AI inputs and outputs.
Key points
- •The model features a built-in guardrail function that evaluates the safety of inputs and outputs.
- •It is designed to detect harmful information within the inputs and outputs of large language models (LLMs).
- •The model is specifically developed to ensure the safe and secure utilization of generative AI.