fix: canonicalize verdict vocabulary (repro-scratio-release-audit)

pages/claim-2-gaussian-accuracy-and-runtime/page.md

@@ -1,50 +1,52 @@
-# Claim 2: Gaussian accuracy and runtime
-
-
----
-<!-- trackio-cell
-{"type": "markdown", "id": "cell_e272c8d3b16e", "created_at": "2026-07-29T20:02:21+00:00", "title": "Claim 2: Gaussian accuracy and runtime"}
--->
-**PARTIAL - 1/2.** The seeded 1,200-step CPU run is finite and gives
-direct/naive MSE `1.6455/5.1616` with
-`1.04x` speedup. Tracked CSV aggregation favors direct ratios
-in all 30 displayed comparisons, with
-cached speedups 1.89x-
-3.09x. Yet each scRatio result row has
-three values while Figure 2 says five runs; raw predictions and paper-scale
-retraining are absent.
-
-
----
-<!-- trackio-cell
-{"type": "code", "id": "cell_1c95b9588ead", "created_at": "2026-07-29T20:02:21+00:00", "title": "C2 evidence", "language": "python"}
--->
-````output
-{
-  "all_direct_faster_than_naive": true,
-  "all_direct_mse_lower_than_naive": true,
-  "baseline_runs_per_configuration": [
-    5
-  ],
-  "cached_rows": 230,
-  "caption_vs_cached_run_count_match": false,
-  "mse_gain_min_median_max": [
-    1.222222222222222,
-    2.2033454823060064,
-    5.995024875621891
-  ],
-  "paper_caption_training_runs": 5,
-  "scratio_cached_times_per_row": [
-    3
-  ],
-  "scratio_cached_values_per_row": [
-    3
-  ],
-  "selected_schedule_comparisons": 30,
-  "speedup_min_median_max": [
-    1.8911516853932582,
-    2.479619540024892,
-    3.0897971822263504
-  ]
-}
-````
+# Claim 2: Gaussian accuracy and runtime
+
+
+---
+<!-- trackio-cell
+{"type": "markdown", "id": "cell_e272c8d3b16e", "created_at": "2026-07-29T20:02:21+00:00", "title": "Claim 2: Gaussian accuracy and runtime"}
+-->
+
+**Verdict: Inconclusive.**
+**PARTIAL - 1/2.** The seeded 1,200-step CPU run is finite and gives
+direct/naive MSE `1.6455/5.1616` with
+`1.04x` speedup. Tracked CSV aggregation favors direct ratios
+in all 30 displayed comparisons, with
+cached speedups 1.89x-
+3.09x. Yet each scRatio result row has
+three values while Figure 2 says five runs; raw predictions and paper-scale
+retraining are absent.
+
+
+---
+<!-- trackio-cell
+{"type": "code", "id": "cell_1c95b9588ead", "created_at": "2026-07-29T20:02:21+00:00", "title": "C2 evidence", "language": "python"}
+-->
+````output
+{
+  "all_direct_faster_than_naive": true,
+  "all_direct_mse_lower_than_naive": true,
+  "baseline_runs_per_configuration": [
+    5
+  ],
+  "cached_rows": 230,
+  "caption_vs_cached_run_count_match": false,
+  "mse_gain_min_median_max": [
+    1.222222222222222,
+    2.2033454823060064,
+    5.995024875621891
+  ],
+  "paper_caption_training_runs": 5,
+  "scratio_cached_times_per_row": [
+    3
+  ],
+  "scratio_cached_values_per_row": [
+    3
+  ],
+  "selected_schedule_comparisons": 30,
+  "speedup_min_median_max": [
+    1.8911516853932582,
+    2.479619540024892,
+    3.0897971822263504
+  ]
+}
+````