Update anova_recompute.py default output path to anova.json

evaluation/anova_recompute.py

@@ -4,7 +4,7 @@ This script is the authoritative source for the F-statistics and partial
 eta-squared values reported in main_v2.tex Findings (line ~341-342).
 
 Usage:
-    # Default (uses evaluation/igf_results.json -> evaluation/significance.json):
+    # Default (uses evaluation/igf_results.json -> evaluation/anova.json):
     python evaluation/anova_recompute.py
 
     # Override input/output paths:
@@ -36,8 +36,8 @@ recommended default in modern statistical practice (R's car package
 default) for designs that are unbalanced or nearly so.
 
 Output:
-  - evaluation/significance.json (additive: does NOT overwrite the
-    legacy `significance.json` so that the two computations are
+  - evaluation/anova.json (additive: does NOT overwrite the
+    legacy `anova.json` so that the two computations are
     transparently reported side-by-side).
 """
 import argparse
@@ -51,7 +51,7 @@ from statsmodels.stats.anova import anova_lm
 
 ROOT = Path(__file__).resolve().parent.parent
 IGF_RESULTS = ROOT / "evaluation" / "igf_results.json"
-OUTPUT = ROOT / "evaluation" / "significance.json"
+OUTPUT = ROOT / "evaluation" / "anova.json"
 
 GEN_MODELS = ["sd15", "sdxl", "sd35", "flux1", "hunyuan", "kolors"]
 GT_KEY = "gt_baseline"
@@ -64,7 +64,7 @@ def parse_args():
     p.add_argument("--input", type=str, default=None,
                    help="Input igf_results.json path (default: evaluation/igf_results.json).")
     p.add_argument("--output", type=str, default=None,
-                   help="Output significance JSON path (default: evaluation/significance.json).")
+                   help="Output significance JSON path (default: evaluation/anova.json).")
     return p.parse_args()