Second

app.py

@@ -4,19 +4,58 @@ import pandas as pd
 # Set page configuration
 st.set_page_config(page_title="Cyber Benchmark Hub: SECQA Leaderboard", layout="wide")
 
-# Main Title (ensures it's displayed on the main page)
+# Main Title
 st.title("Cyber Benchmark Hub: SECQA Leaderboard")
-st.markdown("## Powered by **Priam Cyber AI**")
 st.markdown("#### [View the SECQA Dataset](https://huggingface.co/datasets/zefang-liu/secqa)")
 
+# Sidebar: Logo and Website Link
+with st.sidebar:
+    st.image("https://cdn.prod.website-files.com/630f558f2a15ca1e88a2f774/631f1436ad7a0605fecc5e15_Logo.svg", use_container_width=True)
+    st.markdown("[Priam.ai](https://www.priam.ai/)")
+    st.divider()
+    
+    # Top-level: Dataset Category
+    dataset_categories = ["Multiple Choice", "Open Question", "Steps (Reasoning)"]
+    selected_category = st.selectbox("Select Dataset Category", dataset_categories, index=0)
+    
+    # Filter dataset options based on category
+    datasets_by_category = {
+        "Multiple Choice": ["secQA"],
+        "Open Question": ["Testing..."],
+        "Steps (Reasoning)": ["Testing..."]
+    }
+    dataset_choice = st.selectbox("Select Dataset", datasets_by_category[selected_category], index=0)
+    
+    st.divider()
+    st.header("Filters & Options")
+    dataset_version = st.radio("Select Dataset Version", ["v1", "v2"])
+    # For filtering the leaderboard by model type
+    # Note: The available model types will come from the CSV, once loaded.
+    # We'll load the CSV later and then update this filter accordingly.
+    source_filter_placeholder = st.empty()  # placeholder for source filter after data is loaded
+    
+    st.markdown("---")
+    st.header("Test Parameters")
+    test_params = pd.DataFrame({
+        "Value": [0, 1, 0, 1, 0]
+    }, index=["Temperature", "n", "Presence Penalty", "Top_p", "Frequency Penalty"])
+    st.table(test_params)
+
+# Determine file path based on dataset choice.
+# For now, if dataset_choice is "secQA", we use "Benchmark.csv"
+if dataset_choice == "secQA":
+    file_path = "Benchmark.csv"  # Ensure this file is uploaded in your Hugging Face Space
+else:
+    file_path = "Benchmark.csv"  # Placeholder: update with actual file paths for future datasets
+
 # Function to load and clean CSV data
 @st.cache_data
 def load_data(file_path):
     df = pd.read_csv(file_path)
-
+    
     # Remove any unnamed columns (caused by trailing commas)
     df = df.loc[:, ~df.columns.str.contains('Unnamed', na=False)]
-
+    
     # Standardize column names
     df.columns = df.columns.str.strip()
     df.rename(columns={
@@ -25,35 +64,23 @@ def load_data(file_path):
         "v1 metric": "V1 Accuracy",
         "v2 metric": "V2 Accuracy"
     }, inplace=True)
-
+    
     # Convert percentage strings to floats (e.g., "100%" → 1.0)
     for col in ["V1 Accuracy", "V2 Accuracy"]:
         df[col] = df[col].astype(str).str.replace("%", "").str.strip()
         df[col] = pd.to_numeric(df[col], errors='coerce') / 100
-
+    
     return df
 
 # Load dataset
-file_path = "Benchmark.csv"  # Ensure this file is uploaded in your Hugging Face Space
 df = load_data(file_path)
 
-# Sidebar: Logo, then Filters and Options
-with st.sidebar:
-    st.image("https://cdn.prod.website-files.com/630f558f2a15ca1e88a2f774/631f1436ad7a0605fecc5e15_Logo.svg", use_container_width=True)
-    st.divider()
-    st.header("Filters & Options")
-    dataset_version = st.radio("Select Dataset Version", ["v1", "v2"])
-    source_filter = st.multiselect(
-        "Select Model Type",
-        options=df["Type"].unique().tolist(),
-        default=df["Type"].unique().tolist()
-    )
-    st.markdown("---")
-    st.header("Test Parameters")
-    test_params = pd.DataFrame({
-        "Value": [0, 1, 0, 1, 0]
-    }, index=["Temperature", "n", "Presence Penalty", "Top_p", "Frequency Penalty"])
-    st.table(test_params)
+# Update the source filter with the actual options from the data
+source_filter = source_filter_placeholder.multiselect(
+    "Select Model Type",
+    options=df["Type"].unique().tolist(),
+    default=df["Type"].unique().tolist()
+)
 
 # Apply filtering based on the sidebar selections
 df_filtered = df[df["Type"].isin(source_filter)] if source_filter else df
@@ -70,7 +97,7 @@ df_filtered.insert(0, "Rank", range(1, len(df_filtered) + 1))
 col1, col2 = st.columns([2, 1])
 
 with col1:
-    st.subheader(f"Leaderboard for SECQA Version {dataset_version}")
+    st.subheader(f"Leaderboard for {dataset_choice.upper()} Version {dataset_version}")
     st.dataframe(df_filtered.reset_index(drop=True))
 
 with col2: