Emmimal/llm-eval-layer

A lightweight LLM evaluation layer that turns subjective model outputs into deterministic decisions by scoring attribution, specificity, relevance, and context quality—designed to detect hallucinations and enable production-grade LLM reliability.