文件擷取的有效逐欄位選擇性風險控制：三種失敗模式、有效性階梯，以及條件化何時有價值

摘要

逐欄位接受/審查，且選擇性風險至多為 alpha——唯有在已接受欄位中的錯誤率受控時才接受某欄位——是文件擷取系統所需的信任契約，而自然程序在真實文件上會靜默違反此契約。在來自 800 份 CORD 收據的 13,859 個真實 claude-sonnet-5 欄位（正確率 49.0%）上，我們診斷出三種失敗模式：文件聚類（設計效應 1.84–2.45）、分數重擬洩漏（風險 0.127 時覆蓋率 0.416，在 95% 的分割中違反 alpha=0.10），以及同分聚集病態（退化分數使閾值網格從 0.030 坍縮至 0.001）。我們將修正措施組織為有效性階梯，並按層級說明保證形式。擬合/驗證分割協議可恢復學習式融合的期望選擇性風險控制：在名義 alpha=0.10 下，風險 0.096、覆蓋率 0.318，無容忍帶（生產變體為 0.326）——這是一個平均意義上的操作點，在 47.5% 的重新分割中其實現風險超過 alpha，並非保證書。使用精確二項尾端的 Mondrian Learn-then-Test 可產生每組 PAC 保證書：欄位 iid 為風險 0.068 時覆蓋率 0.171，聚類校正為 0.140，文件 iid 為 0.060——這是唯一與文件特性相符的層級，但誠實而言目前近乎空洞。預先指定的來源分類法 Support-bin 在 sonnet CORD 資料集上贏得所有嚴格度層級（p<1e-4，Bonferroni 校正）——但此勝出在相同文件上使用 haiku 或 qwen 時無法重現——而在較高正確率的語料庫上，合併閾值勝出：條件化恰好有助於合併法無法提供保證之處，而在其他地方則被學習式分數所涵蓋。在未經選擇程序觸及的 claude-haiku-4-5 上進行凍結配置確認，在兩種風險水準下均成立；一項由三位標註者進行的盲式人工金標準稽核驗證實務層級接受集合的風險為 1.3%，相對於其 10% 的預算（Fleiss' kappa=0.83；標籤錯誤單側偏於悲觀）。以 Apache-2.0 授權釋出，並搭載種子固定與回歸測試把關的程序。

English

Per-field accept/review with selective risk at most alpha -- accept a field only if the error rate among accepted fields is controlled -- is the trust contract document-extraction systems need, and the natural procedure silently violates it on real documents. On 13,859 genuine claude-sonnet-5 fields from 800 CORD receipts (49.0% correct) we diagnose three failure modes: document clustering (design effect 1.84-2.45), score-refit leakage (coverage 0.416 at risk 0.127, violating alpha=0.10 in 95% of splits), and a tie-mass pathology (a degenerate score collapses the threshold grid, 0.030 to 0.001). We organize the fixes as a validity ladder, guarantee form stated per tier. A fit/val split protocol restores expected-selective-risk control for a learned fusion: coverage 0.318 at risk 0.096 at nominal alpha=0.10, no tolerance band (production variant 0.326) -- an on-average point whose realized risk exceeds alpha in 47.5% of resplits, not a certificate. Mondrian Learn-then-Test with exact binomial tails yields per-group PAC certificates: field-iid 0.171 at risk 0.068, cluster-corrected 0.140, doc-iid 0.060 -- the only tier matching documents, honestly near-vacuous today. Support-bin, the pre-specified provenance taxonomy, wins every rigor tier on the sonnet CORD capture (p<1e-4, Bonferroni-corrected) -- a win that does not replicate on the same documents under haiku or qwen -- while on higher-accuracy corpora pooled thresholds win: conditioning helps exactly where pooled cannot certify, subsumed by a learned score elsewhere. A frozen-configuration confirmation on selection-untouched claude-haiku-4-5 held at both risk levels, and a blind three-annotator human-gold audit verifies the practical tier's accepted-set risk at 1.3% against its 10% budget (Fleiss' kappa=0.83; labels err one-sidedly pessimistic). Released Apache-2.0 with seed-pinned, regression-gated procedures.