llm-consistency

This paper presents SSAS (Syntactic & Semantic Context Assessment Summarization), a framework designed to improve consistency in LLM-based sentiment prediction by reducing noise and variance through hierarchical classification and iterative summarization. Empirical evaluation on three industry-standard datasets shows up to 30% improvement in data quality and reliability for enterprise decision-making.