Verification
Verification is the process of checking a system, result, or claim against specified requirements, criteria, or evidence to determine whether it conforms or is correct.
Build Financial Software with Generative AI (From Scratch)
Verification is the process of checking generated or modified software for correctness, suitability, and understandability before relying on it.
- How it is used
- Uses verification to assess AI-assisted financial-software code before it applies payment business rules.
- How deep it goes
- Develops a practical workflow through code review, testable design, unit tests, version checks, and independent result checks.
- Examples
- Uses flawed ACH parsers, Type 6 record classification, and a unit-tested IAT type 12 addenda parser.
- What you can do after
- Evaluate/understand generated code by checking domain rules, writing or running focused tests, and independently confirming results.
Read how Build Financial Software with Generative AI (From Scratch) explains verification →
Reinforcement Learning from Human Feedback
A verification function evaluates a candidate answer or behavior against an objectively checkable correctness criterion.
- How it is used
- Provides definitive reward signals for RLVR on tasks whose answers or behavior can be objectively checked.
- How deep it goes
- Explains answer extraction, exact matching, unit-test execution, binary gating, and their role in the RLVR training loop.
- Examples
- Uses a prime-number sum with boxed-answer matching and a Fibonacci implementation checked by unit tests.
- What you can do after
- Apply it via answer matching or unit tests to turn objectively checked model outputs into training rewards.
Read how Reinforcement Learning from Human Feedback explains verification →
AI for Everyday IT
Verification is the process of checking an artifact or result against evidence or requirements to establish its accuracy or validity.
- How it is used
- Uses evidence and human review to prevent inaccurate AI output from entering workplace documents, messages, or decisions.
- How deep it goes
- Explains review routines, audit records, labeling, and email indicators, but stops short of formal verification methods.
- Examples
- Uses nonexistent API endpoints, implausible chatbot advice, AI-processed records, and suspicious SharePoint-style phishing emails.
- What you can do after
- Diagnose questionable AI output and suspicious emails while documenting review responsibility and protecting sensitive information.
Object-Oriented Software Design in C++
A software quality process comprising verification that implementation conforms to specifications and validation that the resulting system satisfies intended needs.
- How it is used
- Uses testing to confirm that implementation fulfills requirements, alongside client validation of the intended application.
- How deep it goes
- Explains functional specifications, black-box tests, unit tests, and regression tests, while noting that tests cannot prove absolute bug freedom.
- Examples
- Uses functional specifications, external black-box test plans, class-level unit tests, and repeated regression tests.
- What you can do after
- Apply it via black-box, unit, and regression testing to check implemented requirements and detect change-induced breakage.
Read how Object-Oriented Software Design in C++ explains verification →
Related concepts
- testincludes · A test is a procedure or check that evaluates behavior using specified inputs or evidence against expected results or criteria.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.