When Relevance Meets Novelty: Dual-Stable Periodic Optimization for Serendipitous Recommendation

Researchers propose Co-Evolutionary Alignment (CoEA), a new recommendation system method that uses dual large language models to balance relevant and novel content suggestions. The system addresses traditional recommendation bias through dynamic optimization that considers both long-term group identity and short-term individual preferences.

• →Traditional recommendation systems create feedback loops that limit user content exploration and cause fatigue.
• →CoEA introduces Dual-Stable Interest Exploration to model both group identity and individual interests simultaneously.
• →Periodic Collaborative Optimization enables dynamic closed-loop optimization using incremental user data.
• →The method uses two LLMs - one for relevance and one for novelty - that continuously fine-tune each other.
• →Extensive experiments demonstrate effectiveness in providing serendipitous recommendations.