Add 'Show all labels' toggle for scatter plots

leaderboard_transformer.py

@@ -970,7 +970,8 @@ def _plot_scatter_plotly(
         agent_col: str = 'Agent',
         name: Optional[str] = None,
         plot_type: str = 'cost',  # 'cost' or 'runtime'
-        mark_by: Optional[str] = None  # 'Company', 'Openness', or 'Country'
+        mark_by: Optional[str] = None,  # 'Company', 'Openness', or 'Country'
+        show_all_labels: bool = False  # Show labels for all points vs only Pareto frontier
 ) -> go.Figure:
     from constants import MARK_BY_DEFAULT
     if mark_by is None:
@@ -1303,57 +1304,69 @@ def _plot_scatter_plotly(
                 layer="above",
             ))
 
-    # --- Section 7: Add Model Name Labels to Frontier Points ---
-    if frontier_rows:
-        frontier_labels_data = []
-        
-        for row in frontier_rows:
+    # --- Section 7: Add Model Name Labels ---
+    # Show labels for all points if show_all_labels is True, otherwise just Pareto frontier
+    if show_all_labels:
+        # Label all data points
+        labels_data = []
+        for _, row in data_plot.iterrows():
             x_val = row[x_col_to_use]
             y_val = row[y_col_to_use]
             
-            # Get the model name for the label
             model_name = row.get('Language Model', '')
             if isinstance(model_name, list):
                 model_name = model_name[0] if model_name else ''
-            # Clean the model name (remove path prefixes)
             model_name = str(model_name).split('/')[-1]
-            # Truncate long names
             if len(model_name) > 25:
                 model_name = model_name[:22] + '...'
             
-            frontier_labels_data.append({
-                'x': x_val,
-                'y': y_val,
-                'label': model_name
-            })
+            labels_data.append({'x': x_val, 'y': y_val, 'label': model_name})
+    elif frontier_rows:
+        # Label only Pareto frontier points
+        labels_data = []
         
-        # Add annotations for each frontier label
-        # For log scale x-axis, annotations need log10(x) coordinates (Plotly issue #2580)
-        for item in frontier_labels_data:
-            x_val = item['x']
-            y_val = item['y']
-            label = item['label']
+        for row in frontier_rows:
+            x_val = row[x_col_to_use]
+            y_val = row[y_col_to_use]
             
-            # Transform x to log10 for annotation positioning on log scale
-            if x_val > 0:
-                x_log = np.log10(x_val)
-            else:
-                x_log = x_min_log
+            model_name = row.get('Language Model', '')
+            if isinstance(model_name, list):
+                model_name = model_name[0] if model_name else ''
+            model_name = str(model_name).split('/')[-1]
+            if len(model_name) > 25:
+                model_name = model_name[:22] + '...'
             
-            fig.add_annotation(
-                x=x_log,
-                y=y_val,
-                text=label,
-                showarrow=False,
-                yshift=25,  # Move label higher above the icon
-                font=dict(
-                    size=10,
-                    color='#0D0D0F',  # neutral-950
-                    family=FONT_FAMILY_SHORT
-                ),
-                xanchor='center',
-                yanchor='bottom'
-            )
+            labels_data.append({'x': x_val, 'y': y_val, 'label': model_name})
+    else:
+        labels_data = []
+    
+    # Add annotations for each label
+    # For log scale x-axis, annotations need log10(x) coordinates (Plotly issue #2580)
+    for item in labels_data:
+        x_val = item['x']
+        y_val = item['y']
+        label = item['label']
+        
+        # Transform x to log10 for annotation positioning on log scale
+        if x_val > 0:
+            x_log = np.log10(x_val)
+        else:
+            x_log = x_min_log
+        
+        fig.add_annotation(
+            x=x_log,
+            y=y_val,
+            text=label,
+            showarrow=False,
+            yshift=25,  # Move label higher above the icon
+            font=dict(
+                size=10,
+                color='#0D0D0F',  # neutral-950
+                family=FONT_FAMILY_SHORT
+            ),
+            xanchor='center',
+            yanchor='bottom'
+        )
 
     # --- Section 8: Configure Layout  ---
     # Use the same axis ranges as calculated for domain coordinates

ui_components.py

@@ -705,7 +705,7 @@ def create_leaderboard_display(
         primary_runtime_col = f"{category_name} Runtime"
     
     # Function to create cost/performance scatter plot from data
-    def create_cost_scatter_plot(df_data, mark_by=MARK_BY_DEFAULT):
+    def create_cost_scatter_plot(df_data, mark_by=MARK_BY_DEFAULT, show_all_labels=False):
         return _plot_scatter_plotly(
             data=df_data,
             x=primary_cost_col if primary_cost_col in df_data.columns else None,
@@ -713,11 +713,12 @@ def create_leaderboard_display(
             agent_col="SDK Version",
             name=category_name,
             plot_type='cost',
-            mark_by=mark_by
+            mark_by=mark_by,
+            show_all_labels=show_all_labels
         )
     
     # Function to create runtime/performance scatter plot from data
-    def create_runtime_scatter_plot(df_data, mark_by=MARK_BY_DEFAULT):
+    def create_runtime_scatter_plot(df_data, mark_by=MARK_BY_DEFAULT, show_all_labels=False):
         return _plot_scatter_plotly(
             data=df_data,
             x=primary_runtime_col if primary_runtime_col in df_data.columns else None,
@@ -725,7 +726,8 @@ def create_leaderboard_display(
             agent_col="SDK Version",
             name=category_name,
             plot_type='runtime',
-            mark_by=mark_by
+            mark_by=mark_by,
+            show_all_labels=show_all_labels
         )
     
     # Create initial cost scatter plots for all filter combinations
@@ -792,6 +794,13 @@ def create_leaderboard_display(
                 )
             else:
                 show_open_only_checkbox = None
+            
+            # Add checkbox for showing all labels on scatter plot
+            show_all_labels_checkbox = gr.Checkbox(
+                label="Show all labels on scatter plots",
+                value=False,
+                elem_id="show-all-labels-toggle"
+            )
         
         with gr.Column(scale=1):
             mark_by_dropdown = gr.Dropdown(
@@ -835,7 +844,7 @@ def create_leaderboard_display(
             )
             
             # Update function for filters - handles checkboxes and mark_by dropdown
-            def update_display(show_incomplete, show_open_only, mark_by):
+            def update_display(show_incomplete, show_open_only, mark_by, show_all_labels):
                 # Determine which dataframe to show based on checkbox states
                 if show_open_only:
                     df_to_show = df_display_open if show_incomplete else df_display_complete_open
@@ -844,9 +853,9 @@ def create_leaderboard_display(
                     df_to_show = df_display_all if show_incomplete else df_display_complete
                     view_df = df_view_full if show_incomplete else df_view_complete
                 
-                # Regenerate plots with current mark_by setting
-                cost_plot = create_cost_scatter_plot(view_df, mark_by)
-                runtime_plot = create_runtime_scatter_plot(view_df, mark_by)
+                # Regenerate plots with current mark_by and show_all_labels settings
+                cost_plot = create_cost_scatter_plot(view_df, mark_by, show_all_labels)
+                runtime_plot = create_runtime_scatter_plot(view_df, mark_by, show_all_labels)
                 return df_to_show, cost_plot, runtime_plot
             
             # Connect checkboxes and dropdown to the update function
@@ -857,6 +866,7 @@ def create_leaderboard_display(
                 # Add a dummy value for show_open_only when checkbox doesn't exist
                 filter_inputs = [show_incomplete_checkbox, gr.State(value=False)]
             filter_inputs.append(mark_by_dropdown)
+            filter_inputs.append(show_all_labels_checkbox)
             
             show_incomplete_checkbox.change(
                 fn=update_display,
@@ -874,6 +884,11 @@ def create_leaderboard_display(
                 inputs=filter_inputs,
                 outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
             )
+            show_all_labels_checkbox.change(
+                fn=update_display,
+                inputs=filter_inputs,
+                outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
+            )
         else:
             dataframe_component = gr.DataFrame(
                 headers=df_headers,
@@ -888,15 +903,15 @@ def create_leaderboard_display(
             )
             
             # Update function for mark_by and optional open_only checkbox
-            def update_display_no_complete(show_open_only, mark_by):
+            def update_display_no_complete(show_open_only, mark_by, show_all_labels):
                 if show_open_only:
                     df_to_show = df_display_open
                     view_df = df_view_open
                 else:
                     df_to_show = df_display_all
                     view_df = df_view_full
-                cost_plot = create_cost_scatter_plot(view_df, mark_by)
-                runtime_plot = create_runtime_scatter_plot(view_df, mark_by)
+                cost_plot = create_cost_scatter_plot(view_df, mark_by, show_all_labels)
+                runtime_plot = create_runtime_scatter_plot(view_df, mark_by, show_all_labels)
                 return df_to_show, cost_plot, runtime_plot
             
             filter_inputs_no_complete = []
@@ -905,6 +920,7 @@ def create_leaderboard_display(
             else:
                 filter_inputs_no_complete.append(gr.State(value=False))
             filter_inputs_no_complete.append(mark_by_dropdown)
+            filter_inputs_no_complete.append(show_all_labels_checkbox)
             
             if show_open_only_checkbox is not None:
                 show_open_only_checkbox.change(
@@ -917,13 +933,18 @@ def create_leaderboard_display(
                 inputs=filter_inputs_no_complete,
                 outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
             )
+            show_all_labels_checkbox.change(
+                fn=update_display_no_complete,
+                inputs=filter_inputs_no_complete,
+                outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
+            )
         
         legend_markdown = create_legend_markdown(category_name)
         gr.HTML(value=legend_markdown, elem_id="legend-markdown")
     
     # Add a timer to periodically check for data updates and refresh the UI
     # This runs every 60 seconds to check if new data is available
-    def check_and_refresh_data(show_incomplete, show_open_only=False, mark_by=MARK_BY_DEFAULT):
+    def check_and_refresh_data(show_incomplete, show_open_only=False, mark_by=MARK_BY_DEFAULT, show_all_labels=False):
         """Check if data has been refreshed and return updated data if so."""
         current_version = get_data_version()
         if current_version > initial_data_version:
@@ -953,16 +974,16 @@ def create_leaderboard_display(
                 new_df_display_open = prepare_df_for_display(new_df_view_open)
                 new_df_display_complete_open = prepare_df_for_display(new_df_view_complete_open)
                 
-                # Create new scatter plots for all combinations (with current mark_by)
-                new_cost_scatter_complete = create_cost_scatter_plot(new_df_view_complete, mark_by) if len(new_df_display_complete) > 0 else go.Figure()
-                new_cost_scatter_all = create_cost_scatter_plot(new_df_view_full, mark_by)
-                new_cost_scatter_open = create_cost_scatter_plot(new_df_view_open, mark_by) if len(new_df_view_open) > 0 else go.Figure()
-                new_cost_scatter_complete_open = create_cost_scatter_plot(new_df_view_complete_open, mark_by) if len(new_df_view_complete_open) > 0 else go.Figure()
+                # Create new scatter plots for all combinations (with current mark_by and show_all_labels)
+                new_cost_scatter_complete = create_cost_scatter_plot(new_df_view_complete, mark_by, show_all_labels) if len(new_df_display_complete) > 0 else go.Figure()
+                new_cost_scatter_all = create_cost_scatter_plot(new_df_view_full, mark_by, show_all_labels)
+                new_cost_scatter_open = create_cost_scatter_plot(new_df_view_open, mark_by, show_all_labels) if len(new_df_view_open) > 0 else go.Figure()
+                new_cost_scatter_complete_open = create_cost_scatter_plot(new_df_view_complete_open, mark_by, show_all_labels) if len(new_df_view_complete_open) > 0 else go.Figure()
                 
-                new_runtime_scatter_complete = create_runtime_scatter_plot(new_df_view_complete, mark_by) if len(new_df_display_complete) > 0 else go.Figure()
-                new_runtime_scatter_all = create_runtime_scatter_plot(new_df_view_full, mark_by)
-                new_runtime_scatter_open = create_runtime_scatter_plot(new_df_view_open, mark_by) if len(new_df_view_open) > 0 else go.Figure()
-                new_runtime_scatter_complete_open = create_runtime_scatter_plot(new_df_view_complete_open, mark_by) if len(new_df_view_complete_open) > 0 else go.Figure()
+                new_runtime_scatter_complete = create_runtime_scatter_plot(new_df_view_complete, mark_by, show_all_labels) if len(new_df_display_complete) > 0 else go.Figure()
+                new_runtime_scatter_all = create_runtime_scatter_plot(new_df_view_full, mark_by, show_all_labels)
+                new_runtime_scatter_open = create_runtime_scatter_plot(new_df_view_open, mark_by, show_all_labels) if len(new_df_view_open) > 0 else go.Figure()
+                new_runtime_scatter_complete_open = create_runtime_scatter_plot(new_df_view_complete_open, mark_by, show_all_labels) if len(new_df_view_complete_open) > 0 else go.Figure()
                 
                 # Return the appropriate data based on checkbox states
                 if show_open_only:
@@ -997,6 +1018,7 @@ def create_leaderboard_display(
         if show_open_only_checkbox is not None:
             timer_inputs.append(show_open_only_checkbox)
         timer_inputs.append(mark_by_dropdown)  # Always include mark_by
+        timer_inputs.append(show_all_labels_checkbox)
         refresh_timer.tick(
             fn=check_and_refresh_data,
             inputs=timer_inputs,
@@ -1004,7 +1026,7 @@ def create_leaderboard_display(
         )
     else:
         # If no incomplete checkbox, always show all data (but still filter by open if needed)
-        def check_and_refresh_all(show_open_only=False, mark_by=MARK_BY_DEFAULT):
+        def check_and_refresh_all(show_open_only=False, mark_by=MARK_BY_DEFAULT, show_all_labels=False):
             current_version = get_data_version()
             if current_version > initial_data_version:
                 print(f"[REFRESH] Data version changed, reloading...")
@@ -1017,8 +1039,8 @@ def create_leaderboard_display(
                         new_df_view_full = new_df_view_full[new_df_view_full['Openness'].str.lower() == 'open'].copy()
                     
                     new_df_display_all = prepare_df_for_display(new_df_view_full)
-                    new_cost_scatter_all = create_cost_scatter_plot(new_df_view_full, mark_by)
-                    new_runtime_scatter_all = create_runtime_scatter_plot(new_df_view_full, mark_by)
+                    new_cost_scatter_all = create_cost_scatter_plot(new_df_view_full, mark_by, show_all_labels)
+                    new_runtime_scatter_all = create_runtime_scatter_plot(new_df_view_full, mark_by, show_all_labels)
                     return new_df_display_all, new_cost_scatter_all, new_runtime_scatter_all
             
             if show_open_only:
@@ -1028,15 +1050,15 @@ def create_leaderboard_display(
         if show_open_only_checkbox is not None:
             refresh_timer.tick(
                 fn=check_and_refresh_all,
-                inputs=[show_open_only_checkbox, mark_by_dropdown],
+                inputs=[show_open_only_checkbox, mark_by_dropdown, show_all_labels_checkbox],
                 outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
             )
         else:
-            def check_and_refresh_simple(mark_by=MARK_BY_DEFAULT):
-                return check_and_refresh_all(False, mark_by)
+            def check_and_refresh_simple(mark_by=MARK_BY_DEFAULT, show_all_labels=False):
+                return check_and_refresh_all(False, mark_by, show_all_labels)
             refresh_timer.tick(
                 fn=check_and_refresh_simple,
-                inputs=[mark_by_dropdown],
+                inputs=[mark_by_dropdown, show_all_labels_checkbox],
                 outputs=[dataframe_component, cost_plot_component, runtime_plot_component]
             )