Leaderboard: per-domain tables and auto-refresh countdown

app.py

@@ -51,23 +51,21 @@ def refresh_leaderboard() -> tuple:
         prepare_values_table(grouped["overall_values"]),
         prepare_ranks_table(grouped["overall_ranks"]),
         build_status_line(status, baseline),
-        REFRESH_SECONDS,
-        f"Next auto-refresh in **{REFRESH_SECONDS}s**",
     ]
     for domain in DOMAINS:
         values_df = grouped["domain_values"].get(domain)
+        ranks_df = grouped["domain_ranks"].get(domain)
         if values_df is not None and not values_df.empty:
             outputs.append(prepare_values_table(values_df))
         else:
             outputs.append(prepare_values_table(grouped["overall_values"].iloc[:0]))
+        if ranks_df is not None and not ranks_df.empty:
+            outputs.append(prepare_ranks_table(ranks_df))
+        else:
+            outputs.append(prepare_ranks_table(grouped["overall_ranks"].iloc[:0]))
     return tuple(outputs)
 
 
-def tick_refresh_progress(remaining: int) -> tuple[int, str]:
-    next_remaining = max(int(remaining) - 1, 0)
-    return next_remaining, f"Next auto-refresh in **{next_remaining}s**"
-
-
 grouped_initial = get_grouped_dfs(RESULTS_PATH)
 DOMAINS: list[str] = grouped_initial["domains"]
 INITIAL_OUTPUTS = refresh_leaderboard()
@@ -78,48 +76,37 @@ with demo:
     gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
     status = gr.Markdown(INITIAL_OUTPUTS[2], elem_classes="markdown-text")
 
-    refresh_progress = gr.Slider(
-        minimum=0,
-        maximum=REFRESH_SECONDS,
-        value=REFRESH_SECONDS,
-        step=1,
-        label="Auto-refresh countdown",
-        interactive=False,
-        elem_id="refresh-progress",
-    )
-    refresh_countdown = gr.Markdown(
-        f"Next auto-refresh in **{REFRESH_SECONDS}s**",
-        elem_classes="markdown-text",
-    )
-
-    with gr.Row():
-        refresh_btn = gr.Button("Refresh now", variant="secondary")
-
-    gr.Markdown("### Overall", elem_classes="domain-heading")
-    overall_values_df = gr.Dataframe(
-        value=INITIAL_OUTPUTS[0],
-        interactive=False,
-        wrap=True,
-        label="Metric values (normalized vs baseline, lower is better)",
-    )
-    overall_ranks_df = gr.Dataframe(
-        value=INITIAL_OUTPUTS[1],
-        interactive=False,
-        wrap=True,
-        label="Ranks (average per-dataset rank, lower is better)",
-    )
-
-    domain_value_dfs: list[gr.Dataframe] = []
-    for i, domain in enumerate(DOMAINS):
-        gr.Markdown(f"### {domain}", elem_classes="domain-heading")
-        domain_value_dfs.append(
-            gr.Dataframe(
-                value=INITIAL_OUTPUTS[5 + i],
+    domain_tables: list[tuple[gr.Dataframe, gr.Dataframe]] = []
+    with gr.Tabs():
+        with gr.Tab("Overall"):
+            overall_values_df = gr.Dataframe(
+                value=INITIAL_OUTPUTS[0],
                 interactive=False,
                 wrap=True,
-                label=f"{domain} · metric values (lower is better)",
+                label="Metric values (normalized vs baseline, lower is better)",
             )
-        )
+            overall_ranks_df = gr.Dataframe(
+                value=INITIAL_OUTPUTS[1],
+                interactive=False,
+                wrap=True,
+                label="Ranks (average per-dataset rank, lower is better)",
+            )
+
+        for i, domain in enumerate(DOMAINS):
+            with gr.Tab(domain):
+                values_df = gr.Dataframe(
+                    value=INITIAL_OUTPUTS[3 + i * 2],
+                    interactive=False,
+                    wrap=True,
+                    label="Metric values (lower is better)",
+                )
+                ranks_df = gr.Dataframe(
+                    value=INITIAL_OUTPUTS[4 + i * 2],
+                    interactive=False,
+                    wrap=True,
+                    label="Ranks (lower is better)",
+                )
+                domain_tables.append((values_df, ranks_df))
 
     with gr.Accordion("About", open=False):
         gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
@@ -132,22 +119,11 @@ with demo:
             show_copy_button=True,
         )
 
-    refresh_outputs = [
-        overall_values_df,
-        overall_ranks_df,
-        status,
-        refresh_progress,
-        refresh_countdown,
-        *domain_value_dfs,
-    ]
+    refresh_outputs: list[gr.Component] = [overall_values_df, overall_ranks_df, status]
+    for values_df, ranks_df in domain_tables:
+        refresh_outputs.extend([values_df, ranks_df])
 
-    refresh_btn.click(refresh_leaderboard, inputs=[], outputs=refresh_outputs)
     demo.load(refresh_leaderboard, inputs=[], outputs=refresh_outputs)
     gr.Timer(value=REFRESH_SECONDS).tick(refresh_leaderboard, inputs=[], outputs=refresh_outputs)
-    gr.Timer(value=1).tick(
-        tick_refresh_progress,
-        inputs=[refresh_progress],
-        outputs=[refresh_progress, refresh_countdown],
-    )
 
 demo.queue(default_concurrency_limit=20).launch()

src/display/css_html_js.py

@@ -2,17 +2,4 @@ custom_css = """
 .markdown-text { font-size: 16px !important; }
 .tab-buttons button { font-size: 20px; }
 #leaderboard-table { margin-top: 15px; }
-.domain-heading {
-    margin-top: 1.75rem;
-    margin-bottom: 0.5rem;
-    font-size: 1.15rem;
-    font-weight: 600;
-}
-#refresh-progress input[type="range"] {
-    accent-color: #f97316;
-}
-#refresh-progress label span {
-    font-size: 0.9rem;
-    color: #64748b;
-}
 """