root/Jyotisha
1
0
Fork 0
Code Issues 1 Pull Requests 1 Actions Packages Projects Releases Wiki Activity
2,770 Commits 64 Branches 0 Tags
bedb4d7bd12d02acea843b260dba1dc8399abfea
Commit Graph
7 Commits
Author SHA1 Message Date
Jesse_ChenandClaude Fable 5.1 fa2e7d17f5 research(rectification): v5 verdict for scoring-method research — LR weights, absence and precision follow-up all fail hard line 1 (BUG-1091 closed_by_design) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
 2026-09-30 00:30:34 +08:00
Jesse_ChenandClaude Fable 5.1 8ceb94da02 research(rectification): holdout v5 — drop current-year events (harness cutoff 2026-09-14), index pages 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
 2026-09-29 19:52:43 +08:00
Jesse_ChenandClaude Fable 5.1 abf2ad6aea research(rectification): scoring-method scaffold — LR weights, absence, precision follow-up pipelines on v4, verdicts pending v5 (BUG-1091) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
 2026-09-29 19:06:45 +08:00
Jesse_ChenandClaude Opus 5.5 0f5442cea2 research(rectification): offline R1/R2/R3 — answer-flip tolerance, V1/V2 rerun, dasha shift arithmetic 
- R1: flipping 1 answer keeps truth in range 98-100% but cuts head hit by
  a third or more; 2 flips squeeze truth out in 7-10% of ±30/±60 replays
  (two flips = 8 points = SEPARATION_LEAD).
- R2: weights do apply (research scorer == production at V0); V1/V2 are
  identity at ±30/±60 by construction and leave six-question metrics
  unchanged at ±10 -> no_benefit (measured). Supplementary V1n does not
  pass the gate.
- R3: boundary shift is ~3.8 days/minute (1.3-5.9), not 1.1; the 45-day
  gate is ~8-34 minutes. The _representative_pairs hypothesis is refuted
  (all-pairs adds no dated probes); the bottleneck is monthly evaluation.
  New finding recorded as BUG-1048 (investigating): _boundary_windows
  year-straddle exemption and positional zip misalignment bypass the gate.
- Dated errata appended (no deletions) to the 09-14/09-16 briefs and
  research docs; README board row -> 待验收. No production code, scoring,
  thresholds, gates or Skill changed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
 2026-09-26 14:47:48 +08:00
jesse-ux 039b0a2608 fix(rectification): say 这两分钟 only when two candidates remain (BUG-692)
Independent Staging Quality Gate / validate (push) Successful in 9m43s
Details
Independent Staging Quality Gate / publish (push) Successful in 8m50s
Details 
Count-aware closed-pool invite, catch the five stale contract assertions, and mark M1b V1/V2 as not_measured.
 2026-09-15 09:43:51 +08:00
Jesse_ChenandClaude Fable 5 62803c5d01 docs: defer tightening the domain gate until the invite ships 
上游 R3 要求淘汰与定案需 ≥3 领域，本仓是 MIN_ACCEPTANCE_DOMAINS=2。产品
2026-09-14 决定先不动，等 BUG-689 的补经历邀请上线、有真机数据后再评估：那时
用户更容易补到第三个领域，收紧代价更小；现在收紧会与「永远给结果」冲突。
写进 BUG-689 单的 §4b 与研究定论页的产品决策存档，并明令执行方不得顺手改常量。

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0155nFCgCHtoA7jhSDGmZmMu
 2026-09-14 14:46:48 +00:00
Jesse_ChenandClaude Fable 5 a643452b44 docs(research): closure page for the two rounds on unseparable candidates 
把 09-14 两轮离线研究和上游对照合并成一页定论：加权重类改法全部不过门、
簇合并从未触发（假线索）、宽度按线上口径是 15/33/56 且真值 20/20 从未被挤出、
问答链有效（回放六题后头名命中 0.80/0.55/0.35）。剩下的唯一痛点是区间内部并列，
只能靠带年月的新经历或诚实呈现。附 KP 计分与风格题前置两条产品决策存档，
以及给下一个人的三条提醒（先确认指标口径、不要从 docstring 推断行为、
真值覆盖率优先于区间宽度）。ACTIVE_FRONTS 加索引。

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0155nFCgCHtoA7jhSDGmZmMu
 2026-09-14 14:18:13 +00:00