import pandas as pd

# Load both datasets
train_df = pd.read_csv("new_dataset.csv")
benchmark_df = pd.read_csv("benchmark_dataset.csv")

# Create a unique string for each row by combining premise and hypothesis
train_pairs = set(train_df['s1'] + " ||| " + train_df['s2'])
benchmark_pairs = set(benchmark_df['s1'] + " ||| " + benchmark_df['s2'])

# Find any exact matches that exist in both sets
leaked_data = train_pairs.intersection(benchmark_pairs)

print(f"Number of leaked samples: {len(leaked_data)}")
if len(leaked_data) > 0:
    print("Warning: You have overlapping data!")