Cognition · Software component

Zero-Shot Step Verifier

Software componentCognitionCognition & Memoryarc:ZeroShotStepVerifier

A reasoning verifier that judges individual reasoning steps by prompting a general-purpose LLM without task-specific training, outputting binary, graded or detailed judgments.

Responsibility. Scores reasoning steps using a prompted general-purpose LLM.

Also known as: Prompted LLM verifier

Variant of Reasoning Verifier abstract

When to choose. Choose when labeled reasoning-correctness data is unavailable and moderate agreement with human judgment suffices, using a properly prompted general LLM.

Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

LLM Inference Service Ch3.6

alternative to variability

Fine-Tuned Step Verifier Ch3.6

Quantitative guidance

As stated by the sources; verify before use.

Zero-shot LLM verification often achieves 70-80% agreement with human judgments (Ch3.6).

Classification

Patterns: LLM-as-verifier
Quality attributes: Maintainability (ISO/IEC 25010)
Risks mitigated: Unverified intermediate steps

Sources

Ch3.6: T. Nguyen, "Trace Analysis and Execution Debugging," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.6. ISBN: 9798244538229.