Verifier-Model Hallucination Detector

Software componentSafety & SecuritySafety, Security & Governancearc:VerifierModelHallucinationDetector

A tool hallucination detector that submits each proposed tool call to a separately trained, calibrated verifier model and allows execution only when the verifier confirms the call is valid.

Responsibility. Validates primary-model tool calls with an independent verifier model's calibrated confidence.

Also known as: Confidence calibration, Two-model architecture, Secondary verifier

Variant of Tool Hallucination Detector abstract

When to choose. Choose when labeled examples of correct and hallucinated tool calls are available to train a well-calibrated verifier, and a second model check per call is acceptable.

Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

Inference Server Ch3.7

alternative to variability

Entropy-Based Hallucination Detector Ch3.7

Design guidance

SHOULD proceed only when the verifier confirms the primary model's tool call makes sense.

Classification

Patterns: Generator-verifier (two-model) architectureConfidence calibration
Quality attributes: Safety (ISO/IEC 25010 | NIST AI RMF: safe)Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)
Risks mitigated: Hallucinated tool invocations passing syntactic checks

Sources

Ch3.7: T. Nguyen, "Tool Usage Auditing," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.7. ISBN: 9798244538229.