LOTUS introduces semantic operators that enhance data processing capabilities by leveraging LLMs in a declarative, efficient, and robust framework.

• Creates a formalized approach to AI-based data transformations with performance guarantees
• Enables complex analytics and reasoning over large knowledge repositories
• Provides superior expressiveness and usability compared to existing row-wise LLM operations
• Demonstrates practical applications including biomedical classification and fact-checking

For the medical sector, LOTUS offers powerful biomedical multi-label classification capabilities, potentially revolutionizing how medical data is processed, categorized, and analyzed at scale while maintaining robustness.

Semantic Operators: A Declarative Model for Rich, AI-based Data Processing