diff --git a/docs/decision-log.md b/docs/decision-log.md index 478f621..180fc54 100644 --- a/docs/decision-log.md +++ b/docs/decision-log.md @@ -447,3 +447,18 @@ 模型的碼),歸檔併入 `runs/02-cluster/`。理由:兩者 之間沒有需要檢視的決策點,拆成兩步只增加讀者要理解 的環節;驗證能力不變,五份產物各自可查。 + +## 2026-08-06 + +- **編碼引述改為每碼一行**:原定每碼引 1–3 行歌詞為證, + 全量實跑時 36 首的輸出遭平台內容過濾攔阻,其中包含 + WAP、God Is A Woman、Super Freaky Girl、Sweetest Pie、 + Kiss Me More 等女性力量語料的核心。診斷非髒話而是歌詞 + 重製量——引述量中位數達歌詞長度的 64%、最高 147%, + 同時違反本研究自訂的「repo 只留分類所引摘錄」規則。 + 改為每碼恰一行後實測該 36 首全數通過,每首標到的碼數 + 不變(中位數 11→12),重製量降至 23%。每碼仍有一行 + 可檢視的歌詞證據,檢視編碼偏差的依據不減。附帶記錄: + 平台輸出過濾非隨機地排除了語料中最核心的一批歌,此 + 現象本身列入論文限制討論。標註步驟已入庫的兩次執行 + 重製量約 43%,處置待議。 diff --git a/docs/run-costs.md b/docs/run-costs.md index 4794177..b7e19c3 100644 --- a/docs/run-costs.md +++ b/docs/run-costs.md @@ -26,8 +26,10 @@ $3/$15、opus-4-6 $5/$25、opus-5 與 fable-5 $10/$50) | 2026-08-05 | 01-02-01-merge | run1 | claude-opus-5 | msgbatch_012SEfebTnjU5uWN6hhfnhhh | 8 分 45 秒 | 63,368 | 64,000 | $1.92 | 已取代(64k 截斷) | | 2026-08-05 | 01-02-01-merge | run1 | claude-opus-5 | msgbatch_01Wi9xEoQKCy1nZK6myVNg77 | 11 分 34 秒 | 63,368 | 70,175 | $2.07 | 已取代(驗證不過:漏 59、發明 140、重複 3) | | 2026-08-05 | 01-02-01-merge | run1 | claude-fable-5 | msgbatch_01M15hUs8P9FDTqWvAHdc1cA | 32 分 45 秒 | 63,368 | 107,721 | $3.01 | 現行歸檔(JSON 語法毀損,驗證不過) | - | 2026-08-05 | 01-02-01-merge | run1 | claude-fable-5 | msgbatch_012stRP28uuQDQwobm41HMLh | — | 0 | 0 | $0.00 | 拒收(thinking.type.enabled 不支援) | | 2026-08-05 | 01-02-01-merge | run1 | claude-fable-5 | msgbatch_01KpS7AMx1zAHTcNujgGsouJ | 26 分 28 秒 | 63,368 | 128,000 | $3.52 | 現行歸檔(effort max:128k 全耗於推理,正文空白) | +| 2026-08-05 | 03-01-code | run1 | claude-sonnet-4-6 | msgbatch_01L7VepTxSNz5vruvzzjuL2c | 6 分 36 秒 | 1,247,264 | 492,730 | $5.57 | 36 首輸出遭內容過濾攔阻,待修訂重跑 | +| 2026-08-05 | 03-01-code | 引述減量實驗(36 首)| claude-sonnet-4-6 | msgbatch_01Hskdhi2DkudYmgZhhgFts7 | 3 分 51 秒 | 56,458 | 11,074 | $0.17 | 實驗:每碼一行引述,36 首全數通過過濾;歸檔不入 repo | +| 2026-08-05 | 03-01-code | run1 | claude-sonnet-4-6 | msgbatch_01GG5Ez9KT1pPwtQaY4sW7tv | 4 分 35 秒 | 1,293,724 | 269,166 | $3.96 | 過濾零攔阻;song-168、song-590 因餘額用盡失敗,song-775 拒答 | -累計支出:$33.70。 +累計支出:$43.40。 diff --git a/prompts/03-01-code.md b/prompts/03-01-code.md index c841bf8..455981b 100644 --- a/prompts/03-01-code.md +++ b/prompts/03-01-code.md @@ -17,9 +17,9 @@ Rules: - Use only the given keywords, spelled exactly as given. - A song may match any number of keywords, including none. -- For each keyword you assign, quote 1 to 3 verbatim lines of - the lyrics that ground it; do not assign a keyword you - cannot ground. +- For each keyword you assign, quote exactly one verbatim + line of the lyrics that grounds it; do not assign a keyword + you cannot ground. Do not wrap the output in a Markdown code fence. The output must be strictly valid JSON; use backslash @@ -30,6 +30,6 @@ its list of quotes (an empty object when no keyword applies), and nothing else: { - "first-keyword": ["quoted line", "another quoted line"], - "second-keyword": ["quoted line"] + "first-keyword": ["quoted line"], + "second-keyword": ["another quoted line"] }