diff --git a/evalharness/cli.py b/evalharness/cli.py
index 6874585..448dee0 100644
--- a/evalharness/cli.py
+++ b/evalharness/cli.py
@@ -262,12 +262,19 @@ def _print_result_panel(console, rep, wall_s: float = 0.0):
     body = Table(show_header=False, box=None, padding=(0, 2))
     body.add_column('k', style='dim', no_wrap=True)
     body.add_column('v', overflow='fold')
+    def score_bar(v, width=22):
+        # rich-Bar-style glyphs with half-cell precision: ━━━╸╌╌
+        filled = max(0.0, min(1.0, v)) * width
+        full = int(filled)
+        half = '╸' if filled - full >= 0.5 else ''
+        return '━' * full + half + '╌' * (width - full - len(half))
+
     for m, v in metrics:
         if m == 'extraction_failure_rate':
             continue
-        bar = Text('█' * int(round(v * 24)) + '·' * (24 - int(round(v * 24))))
-        bar.stylize(color(v))
-        body.add_row(m, Text.assemble(f'{v * 100:6.1f}%  ', bar))
+        body.add_row(m, Text.assemble(
+            (f'{v * 100:6.1f}%  ', f'bold {color(v)}'),
+            (score_bar(v), color(v))))
     stats = []
     stats.append(f'n={rep.num_samples}')
     if wall_s >= 1:
@@ -343,12 +350,19 @@ def _print_result_panel(console, rep, wall_s: float = 0.0):
     body = Table(show_header=False, box=None, padding=(0, 2))
     body.add_column('k', style='dim', no_wrap=True)
     body.add_column('v', overflow='fold')
+    def score_bar(v, width=22):
+        # rich-Bar-style glyphs with half-cell precision: ━━━╸╌╌
+        filled = max(0.0, min(1.0, v)) * width
+        full = int(filled)
+        half = '╸' if filled - full >= 0.5 else ''
+        return '━' * full + half + '╌' * (width - full - len(half))
+
     for m, v in metrics:
         if m == 'extraction_failure_rate':
             continue
-        bar = Text('█' * int(round(v * 24)) + '·' * (24 - int(round(v * 24))))
-        bar.stylize(color(v))
-        body.add_row(m, Text.assemble(f'{v * 100:6.1f}%  ', bar))
+        body.add_row(m, Text.assemble(
+            (f'{v * 100:6.1f}%  ', f'bold {color(v)}'),
+            (score_bar(v), color(v))))
     stats = []
     stats.append(f'n={rep.num_samples}')
     if wall_s >= 1: