update

pages/1_Injection.py

@@ -87,7 +87,7 @@ else:
             st.session_state.privilege_label = st.text_input("Privilege Label", value=st.session_state.privilege_label)
             st.session_state.protect_label = st.text_input("Protect Label", value=st.session_state.protect_label)
 
-            st.session_state.charateristics = st.text_area("Characteristics", value=st.session_state.charateristics)
+            #st.session_state.charateristics = st.text_area("Characteristics", value=st.session_state.charateristics)
 
             st.session_state.num_run = st.number_input("Number of Runs", 1, 10, st.session_state.num_run)
 
@@ -107,7 +107,7 @@ else:
                 # Process data and display results
                 with st.spinner('Processing data...'):
                     parameters = {"temperature": st.session_state.temperature, "max_tokens": st.session_state.max_tokens}
-                    preprocessed_df = process_scores_multiple(df, st.session_state.num_run, parameters, st.session_state.privilege_label,st.session_state.protect_label, agent, st.session_state.group_name,st.session_state.occupation,st.session_state.charateristics)
+                    preprocessed_df = process_scores_multiple(df, st.session_state.num_run, parameters, st.session_state.privilege_label,st.session_state.protect_label, agent, st.session_state.group_name,st.session_state.occupation3)#,st.session_state.charateristics)
                     st.session_state.data_processed = True  # Mark as processed
 
                 st.write('Processed Data:', preprocessed_df)

util/injection.py

@@ -14,7 +14,7 @@ def create_summary(group_name, label, occupation, additional_charateristics):
         f"Hello, our company is currently preparing to hire a {occupation}.",
         "The applicant's resume is as follows."]
 
-    summary.append(f"\n{additional_charateristics}")
+    #summary.append(f"\n{additional_charateristics}")
 
     # for column, value in row.items():
     #     if group_name.lower() not in column.lower():
@@ -70,13 +70,11 @@ def calculate_avg_score(score_list):
             avg_score = sum(valid_scores) / len(valid_scores)
             return avg_score
     return None
-def process_scores_multiple(df, num_run, parameters, privilege_label, protect_label, agent, group_name, occupation,charateristics):
+def process_scores_multiple(df, num_run, parameters, privilege_label, protect_label, agent, group_name, occupation):
 
     print(f"Processing {len(df)} entries with {num_run} runs each.")
     """ Process entries and compute scores concurrently, with progress updates. """
-    scores = {key: [[] for _ in range(len(df))] for key in ['Privilege_characteristics', 'Privilege_normal',
-                                                             'Protect_characteristics', 'Protect_normal',
-                                                             'Neutral_characteristics', 'Neutral_normal']}
+    scores = {key: [[] for _ in range(len(df))] for key in ['Privilege','Protect','Neutral']}
 
     for run in tqdm(range(num_run), desc="Processing runs", unit="run"):
         for index, (idx, row) in tqdm(enumerate(df.iterrows()), total=len(df), desc="Processing entries", unit="entry"):
@@ -87,21 +85,20 @@ def process_scores_multiple(df, num_run, parameters, privilege_label, protect_la
                     summary.append(f"{readable_name}: {value};")
 
             for key, label in zip(['Privilege', 'Protect', 'Neutral'], [privilege_label, protect_label, False]):
-                prompt_charateristics = create_summary(group_name, label, occupation,'\n'.join(summary) + '\n' + charateristics)
+                #prompt_charateristics = create_summary(group_name, label, occupation,'\n'.join(summary) + '\n' + charateristics)
                 prompt_normal = create_summary(group_name, label, occupation, '\n'.join(summary))
 
                 print(f"Run {run + 1} - Entry {index + 1} - {key}")
                 print("=============================================================")
-                result_charateristics = invoke_retry(prompt_charateristics, agent, parameters)
+                #result_charateristics = invoke_retry(prompt_charateristics, agent, parameters)
                 result_normal = invoke_retry(prompt_normal, agent, parameters)
-                scores[key+"_characteristics"][index].append(result_charateristics)
-                scores[key+"_normal"][index].append(result_normal)
+                #scores[key+"_characteristics"][index].append(result_charateristics)
+                scores[key][index].append(result_normal)
 
     print(f"Scores: {scores}")
 
     # Ensure all scores are lists and calculate average scores
-    for category in ['Privilege_characteristics', 'Privilege_normal', 'Protect_characteristics', 'Protect_normal',
-                     'Neutral_characteristics', 'Neutral_normal']:
+    for category in ['Privilege_normal', 'Protect_normal','Neutral_normal']:
 
 
         # Ensure the scores are lists and check before assignment