AI LAB technical archive
Metrics & Scoring
A detailed breakdown of the proprietary metrics used to evaluate AI reliability.
Proprietary Metrics
RAG Grounding Score
Measures the degree to which an answer is strictly supported by the retrieved context. (Range: 0.0 - 1.0)
Reliability Index
A composite metric evaluating tone consistency, intent accuracy, and response stability across 50+ test runs.
Unsafe Response Detection
Identifies toxic content, biased suggestions, and jailbreak attempts using real-time semantic probing.
PII Leak Prevention
Detects sensitive data (Social Security, Credit Cards, etc.) and masks them instantly via the firewall.