What should a sample request specify so reviewers can make a useful decision?

Prepare the following review requirements with your evaluation, domain and data governance roles so the sample leads to a clear decision about suitability and remaining work.

- Name the evaluation decision the sample must support, such as whether the generated evidence is sufficient to distinguish correction selection from document retrieval ability. - Assign a domain reviewer to inspect business logic and an evaluation reviewer to inspect scoring, ensuring neither judgement substitutes for the other during acceptance. - Request a description of grounding inputs and generation boundaries, with enough detail to examine operational plausibility and disclosure risk without requesting real business records. - Specify which artefacts reviewers need to inspect directly, including source relationships, policy applicability notes and explanations of deliberately missing or conflicting evidence. - Define how reviewers will record defects, separating incorrect business logic from unclear language, broken references and disagreements about what the benchmark should measure. - State the conditions that would require rescoping, such as dependence on a downstream process that the requested environment does not represent or expose. - Ask for a manifest connecting the reviewed sample to its schema and task definitions, so acceptance applies to an identifiable artefact rather than a general description.