Update space

app.py

@@ -21,30 +21,18 @@ data = {
 
 df = pd.DataFrame(data)
 
-def format_dataframe(df):
-    # Melt the dataframe to create 'Dataset', 'Metric', and 'Value' columns
-    melted_df = df.melt(id_vars=['Method'], var_name='Dataset_Metric', value_name='Value')
-    melted_df[['Dataset', 'Metric']] = melted_df['Dataset_Metric'].str.split('_', expand=True)
-    melted_df = melted_df.drop('Dataset_Metric', axis=1)
+def format_dataframe(df, dataset):
+    # Filter the dataframe for the selected dataset
+    columns = ['Method'] + [col for col in df.columns if dataset in col]
+    filtered_df = df[columns].copy()
     
-    # Pivot the table to get the desired format
-    pivoted_df = melted_df.pivot_table(values='Value', index=['Method', 'Dataset'], columns='Metric', aggfunc='first').reset_index()
-    pivoted_df = pivoted_df.sort_values(['Dataset', 'MRR'], ascending=[True, False])
+    # Rename columns
+    filtered_df.columns = [col.split('_')[-1] if '_' in col else col for col in filtered_df.columns]
     
-    # Reorder columns
-    column_order = ['Method', 'Dataset', 'Hit@1', 'Hit@5', 'R@20', 'MRR']
-    pivoted_df = pivoted_df.reindex(columns=column_order)
+    # Sort by MRR
+    filtered_df = filtered_df.sort_values('MRR', ascending=False)
     
-    return pivoted_df
-
-def create_leaderboard(dataset):
-    formatted_df = format_dataframe(df)
-    if dataset != 'All':
-        formatted_df = formatted_df[formatted_df['Dataset'] == dataset]
-        formatted_df = formatted_df.drop('Dataset', axis=1)
-    return formatted_df
-
-datasets = ['All'] + sorted(df.columns[1:].str.split('_').str[0].unique().tolist())
+    return filtered_df
 
 css = """
 table > thead {
@@ -61,16 +49,28 @@ table > tbody > tr > td:nth-child(2) > div {
 """
 
 with gr.Blocks(css=css) as demo:
-    gr.Markdown("# STARK-Human(-Generated) Leaderboard")
-    gr.Markdown("Testing results on STARK-Human(-Generated).")
-    
-    with gr.Row():
-        dataset_dropdown = gr.Dropdown(choices=datasets, value='All', label="Select Dataset")
-    
-    leaderboard = gr.DataFrame(interactive=False)
-    
-    dataset_dropdown.change(create_leaderboard, inputs=dataset_dropdown, outputs=leaderboard)
+    gr.Markdown("# Semi-structured Retrieval Benchmark (STaRK) Leaderboard")
+    gr.Markdown("Refer to the [STaRK paper](https://arxiv.org/pdf/2404.13207) for details on metrics, tasks and models.")
     
-    demo.load(create_leaderboard, inputs=dataset_dropdown, outputs=leaderboard)
+    with gr.Tabs() as outer_tabs:
+        with gr.TabItem("Synthesized"):
+            gr.Markdown("## Synthesized Data Results")
+            with gr.Tabs() as inner_tabs_synthesized:
+                with gr.TabItem("Amazon"):
+                    gr.DataFrame(format_dataframe(df, "STARK-AMAZON"))
+                with gr.TabItem("MAG"):
+                    gr.DataFrame(format_dataframe(df, "STARK-MAG"))
+                with gr.TabItem("Prime"):
+                    gr.DataFrame(format_dataframe(df, "STARK-PRIME"))
+        
+        with gr.TabItem("Human-Generated"):
+            gr.Markdown("## Human-Generated Data Results")
+            with gr.Tabs() as inner_tabs_human:
+                with gr.TabItem("Amazon"):
+                    gr.DataFrame(format_dataframe(df, "STARK-AMAZON"))
+                with gr.TabItem("MAG"):
+                    gr.DataFrame(format_dataframe(df, "STARK-MAG"))
+                with gr.TabItem("Prime"):
+                    gr.DataFrame(format_dataframe(df, "STARK-PRIME"))
 
 demo.launch()