Update app.py

app.py

@@ -315,5 +315,39 @@ Important:
         else:
             st.info(f"No results available for {selected_model_details} on {selected_dataset_details}. Please run the evaluation first.")
 
+        st.markdown("---")
+    
+        all_data = []
+        
+        first_model = list(st.session_state.all_results.keys())[0]
+        base_results = st.session_state.all_results[first_model]
+        
+        for question_idx in range(len(base_results)):
+            row = {
+                'dataset': selected_dataset_details, 
+                'question': base_results[question_idx]['question'],
+                'correct_answer': base_results[question_idx]['correct_answer'],
+                'subject': base_results[question_idx]['subject'],
+                'options': ' | '.join(base_results[question_idx]['options'])
+            }
+            
+            for model_name in st.session_state.all_results.keys():
+                model_results = st.session_state.all_results[model_name]
+                row[f'{model_name}_response'] = model_results[question_idx]['model_response']
+                row[f'{model_name}_is_correct'] = model_results[question_idx]['is_correct']
+            
+            all_data.append(row)
+
+        complete_df = pd.DataFrame(all_data)
+        
+        csv = complete_df.to_csv(index=False)
+        
+        st.download_button(
+            label="Download All Results as CSV",
+            data=csv,
+            file_name=f"all_models_{selected_dataset_details}_results.csv",
+            mime="text/csv",
+            key="download_all_results"
+        )
 if __name__ == "__main__":
     main()