reasoning-note arm plumbing

calibration/run.py

@@ -74,6 +74,7 @@ def run_grid(
     rebuild=False,
     strip_schema_legend=False,
     prose_legend=False,
+    reasoning_note=False,
 ):
     """Run every (model, budget) pair through harness.B.launch -- the unmodified
     extended path, only ``reasoning_budget`` varies -- on the operator's questions."""
@@ -89,7 +90,8 @@ def run_grid(
     for index, (model, budget) in enumerate(plan, start=1):
         print(f"=== calibration {index}/{len(plan)}: {model} @ {budget} tokens ===", flush=True)
         variant = (
-            f"{budget}-prose-legend" if prose_legend
+            f"{budget}-reasoning" if reasoning_note
+            else f"{budget}-prose-legend" if prose_legend
             else f"{budget}-no-legend" if strip_schema_legend
             else budget
         )
@@ -113,6 +115,7 @@ def run_grid(
             reasoning_budget=budget,
             question_ids=question_ids,
             strip_schema_legend=strip_schema_legend or prose_legend,
+            reasoning_note=reasoning_note,
             **harness_kwargs,
         )
 
@@ -156,6 +159,11 @@ def main():
         help="legacy-legend arm: prose legend before the code instead of the embedded "
         "schema block; results land in a '<budget>-prose-legend' sibling directory",
     )
+    parser.add_argument(
+        "--reasoning-note", action="store_true", dest="reasoning_note",
+        help="Thinking-with-Spatial-Code step-by-step note prefixed to the "
+        "post-prompt; results land in a '<budget>-reasoning' sibling directory",
+    )
     parser.add_argument("--rebuild", action="store_true")
     args = parser.parse_args()
 
@@ -194,6 +202,7 @@ def main():
             rebuild=args.rebuild,
             strip_schema_legend=args.strip_schema_legend,
             prose_legend=args.prose_legend,
+            reasoning_note=args.reasoning_note,
         )
     except ValueError as exc:
         parser.error(str(exc))