fix: skip non-metrics in benchmark output

Fixes #468

src/heretic/main.py

@@ -1327,6 +1327,15 @@ def run():
                             else:
                                 table.add_column("Value", justify="right")
 
+                            def format_value(value: Any) -> str:
+                                if isinstance(
+                                    value,
+                                    (float, np.floating),
+                                ):
+                                    return f"{value:.4f}"
+                                else:
+                                    return f"{value}"
+
                             try:
                                 first_benchmark = True
 
@@ -1350,22 +1359,16 @@ def run():
                                     first_row = True
 
                                     for metric, value in results.items():
-                                        if metric != "alias":
+                                        # Skip non-metrics.
+                                        if metric in ["name", "alias", "sample_len"]:
+                                            continue
+
                                         if first_row and not first_benchmark:
                                             if benchmark_original_model:
                                                 table.add_row("", "", "", "")
                                             else:
                                                 table.add_row("", "", "")
 
-                                            def format_value(value: Any) -> str:
-                                                if isinstance(
-                                                    value,
-                                                    (float, np.floating),
-                                                ):
-                                                    return f"{value:.4f}"
-                                                else:
-                                                    return f"{value}"
-
                                         cells = [
                                             benchmark.name if first_row else "",
                                             metric,
@@ -1373,9 +1376,7 @@ def run():
                                         ]
                                         if benchmark_original_model:
                                             cells.append(
-                                                    format_value(
+                                                format_value(original_results[metric])
-                                                        original_results[metric]
-                                                    )
                                             )
                                         table.add_row(*cells)