docs(decision): report the external benchmark on the corrected basis

19 scored rather than 20: sku_l06_1.jpg is now the enrolled l06 view, so it is
reported as leaked instead of as a held-out correct accept. 12 correct,
7 abstain, 0 false accepts.

Both leak checks are named in the record and in the external README, since the
enrollment that closed F-12 is exactly the kind of change that silently
converts a held-out row into training data.

recognization/data/external-eval/README.md

@@ -8,25 +8,30 @@ Tier 2.** DECISION.md's Tier 2 is a second labelled store visit; these are
 |---|---|
 | `other-mayo/` | 8 accepted images for `sku_l04`, `sku_l07`, `sku_l09` from 6 source domains, plus the scored result |
 | `l12-reconciliation/` | What `sku_l12` actually is, from the shelf tag and package pixels |
-| `kewpie/` | Kewpie 450 g research (the near-unknown case) |
+| `kewpie/` | 12 accepted images for six SKUs from 6 source domains, plus the scored result. One is now enrolled, so it is excluded from scoring. |
 | `tier3-sources/` | Two reusable pan clips and 11 rejected candidates for a future Tier-3 replay |
 
-## Leak test
+## Leak checks
 
-Every accepted image was template-matched against `add/image6.jpg`. Highest
+Two, both in `code/eval_external.py`, both run before any number is reported:
-score across all 8 is 0.675 (`sku_l07_1`), and the 0.95+ band that identifies a
-crop of the evaluation photo is empty. The images are genuinely external.
 
-## Result: 8 images, 0 false accepts
+- **Against the evaluation photo.** Every image is template-matched against
+  `add/image6.jpg`. The highest score across all 20 is 0.675 (`sku_l07_1`), and
+  the 0.95+ band that identifies a crop of the evaluation photo is empty.
+- **Against the catalog.** An image that has since been enrolled is refused
+  rather than scored. `kewpie/sku_l06_1.jpg` is one: it is now the enrolled l06
+  view, so scoring it would be training accuracy wearing a held-out label.
 
-Scored through the shipped `LocalLaneAdapter` + `DecisionPolicy` against
+## Result: 19 scored, 0 false accepts
-`stock-lopia-clean` (`other-mayo/eval_result.json`):
 
-| Verdict | Count | Files |
+    .venv/bin/python code/eval_external.py --stock stock-lopia-clean \
-|---|---|---|
+        data/external-eval/kewpie data/external-eval/other-mayo
-| Correct accept | 4 | `sku_l07_1`, `sku_l07_2`, `sku_l09_1`, `sku_l09_2` |
+
-| Abstain | 4 | all `sku_l04_*` |
+| Set | Scored | Correct | Abstain | False accepts | Leaked |
-| False accept | **0** | — |
+|---|---|---|---|---|---|
+| `kewpie/` | 11 | 8 | 3 | **0** | 1 |
+| `other-mayo/` | 8 | 4 | 4 | **0** | 0 |
+| **Total** | **19** | **12** | **7** | **0** | **1** |
 
 Two things this establishes that the image6 shelf cannot:
 

recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md

@@ -835,7 +835,7 @@ out of "210g"), and percentages are excluded because "カロリー 50% カット
 claim about the contents. Measured on the corrected labels, full resolution:
 precision 86% -> 91%, false accepts 8 -> 5, coverage unchanged at 79%.
 
-### External benchmark: 20 images, 0 false accepts
+### External benchmark: 19 scored, 0 false accepts
 
 The image6 shelf is one photo with one label set, so every number above is
 measured on data the design was tuned against. Twenty images for eight SKUs
@@ -843,17 +843,25 @@ were collected from independent sources (six retailer and marketplace domains)
 to test the pipeline on data it had no part in building. None is Tier 2; all
 are `external_proxy` and every manifest says so.
 
-Leak-tested first: the highest template match of any of the 20 against
+Two leak checks run before any number is reported, both in
-`add/image6.jpg` is 0.675, so none is a crop of the evaluation photo.
+`code/eval_external.py`:
+
+- Against the evaluation photo: the highest template match of any image against
+  `add/image6.jpg` is 0.675, so none is a crop of it.
+- Against the catalog: an image that has since been enrolled is refused rather
+  than scored. `sku_l06_1.jpg` is one - it is now the enrolled l06 view, so
+  scoring it would be training accuracy wearing a held-out label. It is
+  reported as `leaked`, not as a correct accept.
 
 Scored through the shipped `LocalLaneAdapter` + `DecisionPolicy` against
 `stock-lopia-clean` (`data/external-eval/*/eval_result.json`):
 
-| Set | Images | Correct | Abstain | False accepts |
+| Set | Scored | Correct | Abstain | False accepts | Leaked |
-|---|---|---|---|---|
+|---|---|---|---|---|---|
-| `kewpie/` | 12 | 9 | 3 | **0** |
+| `kewpie/` | 11 | 8 | 3 | **0** | 1 |
-| `other-mayo/` | 8 | 4 | 4 | **0** |
+| `other-mayo/` | 8 | 4 | 4 | **0** | 0 |
-| **Total** | **20** | **13** | **7** | **0** |
+| **Total** | **19** | **12** | **7** | **0** | **1** |
+
 - **`sku_l07` is answered with no enrolled image.** Both its images accept
   through the text lane alone - one reads ラッキー フヨネーズ, the other the
   rear label's 製造者：丸和油脂株式会社 with 内容量：200g, which together