Key answer
Existence is only the first check; assess continued applicability and whether the answer broadens the original conclusion. Older is not automatically obsolete and newer not automatically reliable.
Full explanation
Focus on three questions:
· Matching time/version? For features, prices and policies, check current official documentation, effective dates, versions and account types. For papers, look for revisions, corrections and retractions; Crossmark can show publisher-registered updates. Crossref documentation
· Matching context? Read surrounding passages, methods and results for omitted negations, conditions, comparators and exceptions. Watch correlation becoming causation, improvement on one metric becoming overall superiority, and relative changes becoming absolute changes.
· Matching subjects? Compare populations/models, tasks, interventions, baselines and metrics. Cochrane treats mismatches in populations, interventions, comparisons and outcomes as indirectness. Adapt those dimensions to AI models/tasks. Handbook Section 14.2.2 and Table 14.2.b
A prompt improving English multiple-choice accuracy does not establish better Chinese writing universally; language, task and metric all changed. State the measured result and label transfer to new uses as requiring testing.
Check the conclusion that this method suits my task. List source date/version, subjects, comparators and metrics, and differences from my setting. Retain important conditions and explain revisions or conflicting findings.
Prioritize differences affecting decisions rather than fully reviewing every paper. An abstract supports its public overview; experimental-detail judgments need full text or supplementary material.
References
1. Cochrane, Handbook for Systematic Reviews of Interventions, Section 14.2.2 and Table 14.2.b
2. Crossref, Crossmark documentation, updates, corrections and retraction status