File size: 4,396 Bytes
45e320b 6037f36 24a0949 6037f36 45e320b 6037f36 45e320b 6037f36 45e320b 6037f36 45e320b 6037f36 45e320b 6037f36 45e320b 24a0949 45e320b 24a0949 45e320b 6037f36 24a0949 45e320b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 |
import streamlit as st
import pandas as pd
st.set_page_config(page_title="Cyber Benchmark Hub: SECQA Leaderboard", layout="wide")
st.title("Cyber Benchmark Hub: SECQA Leaderboard")
st.markdown("#### [View the SECQA Dataset](https://huggingface.co/datasets/zefang-liu/secqa)")
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()
dataset_categories = ["Multiple Choice"]
selected_category = st.selectbox("Select Dataset Category", dataset_categories, index=0)
datasets_by_category = {
"Multiple Choice": ["secQA"],
}
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={
"model name": "Model",
"source": "Type",
"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
df = load_data(file_path)
# 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
# Choose the correct metric version and compute Accuracy
df_filtered["Accuracy"] = df_filtered["V1 Accuracy"] if dataset_version == "v1" else df_filtered["V2 Accuracy"]
df_filtered = df_filtered[["Model", "Type", "Accuracy"]].dropna() # Drop rows with errors
# Sort by Accuracy descending
df_filtered = df_filtered.sort_values("Accuracy", ascending=False).reset_index(drop=True)
# Compute dense ranking so that models with equal accuracy share the same rank
df_filtered['Rank'] = df_filtered['Accuracy'].rank(method='dense', ascending=False).astype(int)
df_filtered = df_filtered[['Rank', 'Model', 'Type', 'Accuracy']]
# Use columns to display leaderboard and model details side-by-side
col1, col2 = st.columns([2, 1])
with col1:
st.subheader(f"Leaderboard for {dataset_choice.upper()} Version {dataset_version}")
st.dataframe(df_filtered.style.hide(axis='index'))
with col2:
st.subheader("Model Details")
selected_model = st.selectbox("Select a Model", df_filtered["Model"].tolist())
model_details = df_filtered[df_filtered["Model"] == selected_model].iloc[0]
st.write(f"**Model:** {model_details['Model']}")
st.write(f"**Type:** {model_details['Type']}")
st.write(f"**Accuracy:** {model_details['Accuracy']:.2%}")
st.write(f"**Rank:** {model_details['Rank']}")
# Footer
st.markdown("---")
st.info("More dataset benchmarks will be added to this hub in the future.")
|