2 commits

Author	SHA1	Message	Date
wangzhuc	792aec4101	refine: optimize exemplar seeds with data-driven segment selection ... Replace 5 seed segments sourced from weaker articles (score 47-57) with segments from the highest-scoring articles (score 32-37): - Opening: add 国内+日本留学 label-contrast opening (stddev 45.7) - Opening: add 长期主义骗局 belief-then-shatter opening - Emotional: add 国内+日本留学 温水煮青蛙 stagnation segment - Closing: replace 人夫约会指南 (57.1) with 庞氏骗局 (36.7) - Closing: replace 天才在左 (51.7) with 穷人没教育 (33.0) All replacements maintain pattern diversity while upgrading to segments from data-proven most-human articles. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-03-30 23:32:10 +08:00
wangzhuc	d2b87c489c	feat: add universal exemplar seeds for users without articles ... Seeds demonstrate anti-AI structural patterns (sentence variance, real negative emotion, self-correction, abrupt closings) without imposing a specific writing style. Step 4.4 falls back to seeds when the user's exemplar library is empty. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-03-30 22:53:37 +08:00