A. (에이닷) persona dialogue

April 2023 ~ December 2023

My Contributions

Developing and validating the A. Personalization Model

• Designed formatted responses for personalized conversations, referencing the alignment training methodology of the LIMA.
• Managed outsourcing to collect 1k personalized conversation data samples.
• Activated personalized conversation capabilities through LLM instruction tuning.
• Achieved a 20.8% improvement across four persona grounding metrics and reached an 82.5% win rate in qualitative evaluation.
• Collaborated with the RAG system development team to align the LLM serving interface.
• Conducted prompt engineering to enhance RAG system performance.