Fable
The Hanlin Academy had two kinds of compilation output: one ranked candidates and posted results, the other set the standard answer
The Hanlin Academy compiled two sets of documents each spring and autumn. The spring compilation was called the "Examination Question Book"—it contained twenty sets of classical-exegesis questions and ten sets of policy-discussion questions, each with scoring points and grading standards, and after the exam the rankings of students from each province were published by total score. The autumn compilation was called the "True Solutions Collection"—from the excellent answer scripts solicited from each province, the Hanlin Academy selected one hundred pieces, each annotated separately by three editors, and under the answer with the greatest consensus drew a red line, designating it as the standard solution for that question.
The examination question book's role is very clear: it ranks candidates' abilities horizontally. One third of the questions used this year must be replaced next year, because if used questions are repeated, candidates will prepare in advance and scores will not reflect true ability. The scoring points are also fine-tuned each year, introducing new question types and new scoring dimensions. Its value lies in comparison and ranking.
The logic of the ground-truth solution set is completely different. It does not rank—it simply pairs each of one hundred questions with a standard answer. The value of this answer set is not telling test-takers their rank; the key is providing an anchor for future grading—when a grader hesitates over a new answer sheet's score, they can place it alongside the standard answers in the ground-truth set for comparison. The standard answers are fixed until enough evidence proves they need updating.
After this dual-track system ran for many years, a management disaster occurred. The new chief editor thought the examination question book and the true answer collection were too similar—both were collections of questions plus answers—and proposed merging them into one "Imperial Academy Compendium," with each question having both scoring criteria and model answers, under unified management. The merger initially seemed convenient, but a year later problems arose.
The standard answers to some questions in the exam question book were leaked by candidates, who memorized the answers to the original questions from the true solution set at their prep institutions. That year's exam rankings were severely distorted—high scores were not the result of high ability, but of good memory. The chief editor was forced to restore the dual-track system and added a new rule: the source of questions in the exam question book must not overlap with the source of questions in the true solution set. If the true solution set has already been made public, the exam question book cannot use questions from it.
After restoring the dual-track system, the Hanlin Academy added one more time boundary: the true solution set could be given to editors and examiners for routine calibration of standards, while the examination question book could only be unsealed on the day of the exam. The former was an anchor, helping everyone know what a good answer looked like; the latter was the arena, responsible for comparing who could truly write.
From then on, the editors no longer said 'it's all just questions plus answers anyway.' The question booklet was used for final ranking, and the true solution set for daily calibration; one had to be protected from leaks, the other had to stay stable. Mixing the two looked like less trouble in the short term, but in the long term it would taint every high score with the suspicion of having seen the questions in advance.