How should duplicates be treated in evaluation?

• A. Always equal since duplicates exist.
• B. Treat duplicates as separate results and rate independently.
• C. Favor the most relevant, authoritative version and deprioritize duplicates.
• D. Delete duplicates immediately.

Answer: (C)

Duplicates can dilute the usefulness of evaluation results, so the goal is to surface the best quality version. The best approach is to favor the most relevant, authoritative version and deprioritize duplicates. Determine relevance by how well the content matches the user's intent, how current it is, and the credibility of the source; the top result should come from the most authoritative source, with other copies ranked lower or presented after the best one. This keeps results accurate and streamlined rather than cluttered by near-identical content. Deeper reasons: treating duplicates as equal ignores differences in quality, listing duplicates as separate results wastes space and confuses users, and deleting duplicates outright can erase provenance; deprioritizing preserves context while maintaining overall result quality.