Flag 4 items with multiplier/'whichever higher' fine caps as ambiguous

README.md

@@ -54,7 +54,15 @@ generator and from kd-4b); each disagreement was adjudicated by `gpt-5.6-sol`. T
 
 As an anchoring check, the 25 disagreements were also re-solved by `gpt-5.6-sol` *without* the code label: 25/25 matched the
 code label. The one ambiguous item (`t1-ef05045ce360`) states a fine as a multiple of an unstated compensation amount, so
-"is the cap exceeded?" has no single answer; it is kept and flagged. Upper bounds are two-sided 95% Clopper-Pearson over
+"is the cap exceeded?" has no single answer.
+
+**Correction (2026-09-26, rule scan after the audit).** Scanning the line each question targets for multiplier or
+"whichever is higher" caps (`N배`, `높은 금액`, `초과 금액`) found **4** such items, all in lawset-v2 (3 calibration, 1 shift):
+`t1-777a210262b4`, `t1-9c41a700acfc`, `t3-d06216b43e5e`, `t1-ef05045ce360`. The LLM audit flagged only the last one; for the
+other three it accepted the code label, which reads only the fixed amount. Treat all four as ambiguous (listed in
+`flags/variable_cap.json`); the released files are unchanged so their hashes stay valid. Removing them moves the lawset-v2
+α = 0.10 result from 18.1% abstained / 2.5% error to 19.0% / 1.8%; lawset-v3 and decree-e2 contain none. The generator now
+excludes these forms for future versions. So the audit bound above is about **wrong** labels; the ambiguous rate is 4 / 2,812. Upper bounds are two-sided 95% Clopper-Pearson over
 decided items. Cost: about 1.01M input / 0.14M output tokens on luna and ~20k tokens on sol.
 
 ## Reference result (kd-4b-ko-v0, Learn-then-Test, δ = 0.10, per decision type)

flags/variable_cap.json

@@ -0,0 +1,103 @@
+{
+ "rule": "kdca.lawgen.VAR_CAP on the line the question targets (kdca.lawgen.VAR_CAP: N배 | 높은 금액 | 초과 금액)",
+ "ambiguous_ids": [
+  "t1-777a210262b4",
+  "t1-9c41a700acfc",
+  "t1-ef05045ce360",
+  "t3-d06216b43e5e"
+ ],
+ "sensitivity": {
+  "lawset-v2": {
+   "as_released": {
+    "0.05": {
+     "abstain_rate": 0.5959,
+     "answered": 219,
+     "selective_error": 0.0
+    },
+    "0.1": {
+     "abstain_rate": 0.1808,
+     "answered": 444,
+     "selective_error": 0.0248
+    },
+    "0.15": {
+     "abstain_rate": 0.1458,
+     "answered": 463,
+     "selective_error": 0.0367
+    },
+    "0.2": {
+     "abstain_rate": 0.0923,
+     "answered": 492,
+     "selective_error": 0.0508
+    }
+   },
+   "flagged_removed": {
+    "0.05": {
+     "abstain_rate": 0.5959,
+     "answered": 219,
+     "selective_error": 0.0
+    },
+    "0.1": {
+     "abstain_rate": 0.19,
+     "answered": 439,
+     "selective_error": 0.0182
+    },
+    "0.15": {
+     "abstain_rate": 0.155,
+     "answered": 458,
+     "selective_error": 0.0306
+    },
+    "0.2": {
+     "abstain_rate": 0.1052,
+     "answered": 485,
+     "selective_error": 0.0474
+    }
+   }
+  },
+  "lawset-v3": {
+   "as_released": {
+    "0.05": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.1": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.15": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.2": {
+     "abstain_rate": 0.1497,
+     "answered": 250,
+     "selective_error": 0.028
+    }
+   },
+   "flagged_removed": {
+    "0.05": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.1": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.15": {
+     "abstain_rate": 0.3401,
+     "answered": 194,
+     "selective_error": 0.0
+    },
+    "0.2": {
+     "abstain_rate": 0.1497,
+     "answered": 250,
+     "selective_error": 0.028
+    }
+   }
+  }
+ }
+}