Agentic AI Reference Architecture

Safety & Security · Software component

Self-Check LLM Fact Checker

Software componentSafety & SecuritySafety, Security & Governancearc:SelfCheckLLMFactChecker

A fact checking rail that runs the LLM a second time to critique its own output against sources, catching subtle logical inconsistencies at roughly double inference cost.

Responsibility. Verifies reasoning-heavy responses by LLM self-critique.

Also known as: Self-check LLM

Variant of Fact Checking Rail abstract

When to choose. Choose for reasoning-heavy domains such as medical diagnosis where semantic similarity alone misses logical requirements.

invokesspecializesis target of alternativeTois invoked byis target of alternativeToLLM Inference Service: invokesLLM Inference ServiceFact Checking Rail: specializesFact Checking RailAlignment-Score Fact Checker: is target of alternativeToAlignment-Score Fact Che…Cascaded Fact Checker: is invoked byCascaded Fact CheckerNLI Fact Checker: is target of alternativeToNLI Fact Checker
Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

is invoked by dependency

alternative to variability

Quantitative guidance

As stated by the sources; verify before use.

Classification

Patterns
LLM self-critique
Quality attributes
Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)

Sources

  1. Ch7.1A: T. Nguyen, "Advanced Implementation with Nvidia NEMO Framework and Nvlink," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 7.1A. ISBN: 9798244538229.