diff --git a/docs/tasks/PROGRESS-rectification-holdout-expansion-20260929.md b/docs/tasks/PROGRESS-rectification-holdout-expansion-20260929.md
index f6e7e21d..f1e0ca26 100644
--- a/docs/tasks/PROGRESS-rectification-holdout-expansion-20260929.md
+++ b/docs/tasks/PROGRESS-rectification-holdout-expansion-20260929.md
@@ -39,6 +39,7 @@ v5 已建：**77 例**（v4 20 例原样继承 + 57 例新增公开 Rodden-AA）
 - `scripts/research/holdout_v5_baseline.py`：把 `minute_resolution_sweep` / `cluster_width_probe` / `futile_collect_stop_replay`（经 `guided_collect_holdout_replay.load_cases`）的模块常量指到指定数据集与独立输出文件，只跑 `baseline` 变体；不改这三个脚本。
 - **硬红线 5**：`--dataset v4` 复跑，三个脚本共 7+12+6 组数字与 `minute_resolution_2026_09_14.json`、`cluster_width_2026_09_14.json`、`futile_collect_stop_replay_2026_09_29.json` 逐格一致（表在 `holdout_v5_build_2026_09_29.md` §3；JSON `holdout_v5_baseline_v4_all_2026_09_29.json`）。
 - v5 全集：`holdout_v5_baseline_v5_all_2026_09_29.json`，数字见构建说明 §4。
+- 发现并处理：`cluster_width_probe.run` 会顺手重写 `minute_resolution_2026_09_14.md` 里的勘误表（数值相同、格式 `15.0`→`15`）。已 `git checkout` 还原该文件，并在 `holdout_v5_baseline.py` 里把 `append_previous_errata` 置空，保证复跑不碰已发布文档。
 
 | 半径 | 引擎先验头名（v4） | 六题回放头名（v4） | 六题后真值在区间（v4） | 六题后宽度中位（v4） | D1 六格真值在区间 truth / opposite |
 | --- | --- | --- | --- | --- | --- |
diff --git a/scripts/research/holdout_v5_baseline.py b/scripts/research/holdout_v5_baseline.py
index 502ef84c..2cfb0068 100644
--- a/scripts/research/holdout_v5_baseline.py
+++ b/scripts/research/holdout_v5_baseline.py
@@ -66,6 +66,11 @@ def baseline_only(module: Any) -> None:
     module.all_variants = lambda: [module.Variant("baseline")]
 
 
+def keep_published_reports_untouched() -> None:
+    """cluster_width_probe.run rewrites the errata table inside the 09-14 sweep report; never do that here."""
+    cwp.append_previous_errata = lambda *args, **kwargs: None
+
+
 def run_sweep(dataset: Path, out_dir: Path, tag: str, radii: tuple[int, ...]) -> dict[str, Any]:
     mrs.HOLDOUT = dataset
     mrs.REPORT_MD = out_dir / f"holdout_v5_baseline_sweep_{tag}.md"
@@ -80,6 +85,7 @@ def run_cluster_width(dataset: Path, out_dir: Path, tag: str, radii: tuple[int,
     cwp.REPORT_MD = out_dir / f"holdout_v5_baseline_cluster_width_{tag}.md"
     cwp.REPORT_JSON = out_dir / f"holdout_v5_baseline_cluster_width_{tag}.json"
     baseline_only(cwp)
+    keep_published_reports_untouched()
     payload = cwp.run(argparse.Namespace(limit=0, radii=[str(r) for r in radii]))
     replay = {
         str(radius): payload["by_radius"][str(radius)]["baseline"].get("replay")