Based on Array

Start by saying that an array-based label does not determine a unique solution. Ask for a sample input and expected output, whether elements are sorted, whether duplicates or negative values appear, and whether output needs original indices. From there, derive an invariant rather than reaching for a pattern by name. A duplicate-removal scan maintains a valid compacted prefix; a subarray-sum approach maintains prior prefix totals; a sorted-pair search maintains that every discarded pair is impossible. Explain why the invariant survives one loop iteration and what condition ends the loop. Test size zero and one, all identical values, extremes that can overflow arithmetic, and results located at boundaries. If memory is constrained, explicitly compare an in-place scan with a hash-based alternative. A single pass is O(n) time and O(1) additional space, while a lookup map is O(n) expected time and O(n) space. This is more reliable than fabricating a problem from a partial report.