11 Commits

Include renames

Author	SHA1	Message	Date
jesse-ux	507ef959f5	research(jev-intent): fix2 把来源 B 现行与高置信错误补进报告 ... 离线从 cache 聚合，不重跑模型。无焦点层 Flash 69.7% 低于 Jev 78.8%。采集层相对门槛标不可判。	2026-09-19 12:28:43 +08:00
jesse-ux	c2ecbc74cd	research(jev-intent): 修复轮重造语料并全量对照 ... 来源 C 改为 DeepSeek Flash 生成+独立复核，撤回模板拼接结论。来源 B 157 条人工标注后跑 Jev x2 与 Flash 全量对照，结论为缺数据。	2026-09-19 11:30:31 +08:00
jesse-ux	d6fc4fb8b3	research(rectification): DeepSeek Flash 抽 33% 对照现行分类器提示词 ... 来源 B 仍为 0。Flash 套生产提示词，采集题相对门槛未过。结论仍不可接。	2026-09-19 09:45:29 +08:00
jesse-ux	fde541c2ca	research(rectification): Jev 意图分类离线对照，结论不可接 ... 来源 C 900 条 + jev-1.13.0 双跑。高置信错误 7%、点选题 answer_class 65%。不改线上分类器。	2026-09-19 09:27:10 +08:00
Jesse_ChenandClaude Fable 5.1	dc732825d8	fix(rectification): 有题就接着问，题问完才出卡；引导窗口不再硬贴领域 ... 出卡时机只看题源有没有空：撤回「门槛达标就短路采集线」的写法，同时 按 D2 保住「题源全空就按现行规则出卡」——门槛只在还有题可问时挡住 出卡，precision_gate_met 改成只上报（新挂在决策与公开投影上），不再 单独决定时机。引导窗口题在无领域轨道上改问开放题，一个时间窗只问一 次；录入卡提交的是「YYYY 年 M 月，<领域>方面有一件事」，不再是题干 的三选一列表。记忆化 golden 只补一个新键并冻结墙钟。离线回放改成注 入真值方向的边界事件，另跑一组反方向对照。Skill 10.0.28。 BUG-747～752 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JUei7K13cYxLHE3Axe4A45	2026-09-16 12:16:55 +00:00
jesse-ux	cfb41daf3d	feat(rectification): 出卡加精度门槛，补经历改成系统点名 ... 宽度超过 10 分钟或头名并列时不再出交付卡，改为按大运边界逐条问、 用类型芯片和年/月选择器录入。跳过的线换问法再问一次；答「这类事 都没有过」的不再问。用户说「没有了」仍立刻给目前范围。Skill 10.0.27。 BUG-740～743	2026-09-16 18:35:27 +08:00
jesse-ux	dfb572fc7f	docs(research): measure precision-adaptive probe gates; no variant clears the bar ... 20 public AA cases, raman/mean. Lowering MIN_BOUNDARY_DAYS narrows ±10 from 15 to 11 minutes but drops top-1 0.80→0.75; wider radii get wider ranges. Varga sensitivity weights (V1/V2) match production; D60 (V3) hurts ±10. Day-precision events offset ±7 never squeeze the true minute out. Production 45/30 gate and equal varga weights stay unchanged.	2026-09-15 09:04:03 +08:00
jesse-ux	cf972f4040	docs(research): measure why delivered range width equals the search window ... Offline v4 holdout probe. Last round's width=window result was the no-elimination metric; production still-valid ranges after six answers are 15/33/56 minutes. Adjacent merge never fires under step-2 radii. W1/W2 match baseline; W3 is uncertain after one coverage squeeze. No production clustering or scoring defaults changed.	2026-09-14 22:11:47 +08:00
jesse-ux	2d2467dca1	docs(research): measure minute-resolution scoring; no variant clears all radii ... Build holdout v4 from the public AA set, correct the two v3 dates, and sweep R1–R5 plus pairs offline. No production scoring defaults change. No implementation brief: delivered width stays the full window on every radius.	2026-09-14 20:43:27 +08:00
jesse-ux	53565d0da1	docs(research): measure house-lord gochara rules with no benefit ... Offline H1-H4 measurement on 20 public AA holdout cases. Block unique top-1 did not rise; H4 made it worse; minute layer unchanged. Leave production scoring untouched. Transits must not drive minute conclusions.	2026-09-14 01:07:11 +08:00
Jesse_ChenandCursor	b063c66835	docs(research): measure dated probe supply after six answers ... Offline holdout replay shows refresh-only R3+R4 add discriminative dated probes; R1/R2 do not meet the gate. Co-authored-by: Cursor <cursoragent@cursor.com>	2026-09-13 12:59:46 +08:00