Update app.py

app.py

@@ -29,7 +29,7 @@ custom_css = """
 
 # Define the columns for the different leaderboards
 UGI_COLS = ['#P', 'Model', 'UGI 🏆', 'W/10 👍', 'Unruly', 'Internet', 'Stats', 'Writing', 'PolContro']
-WRITING_STYLE_COLS = ['#P', 'Model', 'RegV1 🏆', 'RegV2 🏆', 'MyScore 🏆', 'ASSS⬇️', 'SMOG⬆️', 'Yule⬇️']
+WRITING_STYLE_COLS = ['#P', 'Model', 'Reg+MyScore 🏆', 'Reg 🏆', 'MyScore 🏆', 'ASSS⬇️', 'SMOG⬆️', 'Yule⬇️']
 ANIME_RATING_COLS = ['#P', 'Model', 'Score 🏆', 'Dif', 'Cor', 'Std']
 
 # Load the leaderboard data from a CSV file
@@ -147,7 +147,7 @@ with GraInter:
             """)
 
         with gr.TabItem("Writing Style"):
-            leaderboard_df_ws = leaderboard_df.sort_values(by='RegV1 🏆', ascending=False)
+            leaderboard_df_ws = leaderboard_df.sort_values(by='Reg+MyScore 🏆', ascending=False)
             datatypes_ws = ['html' if col == 'Model' else 'str' for col in WRITING_STYLE_COLS]
             leaderboard_table_ws = gr.Dataframe(
                 value=leaderboard_df_ws[WRITING_STYLE_COLS],
@@ -158,7 +158,7 @@ with GraInter:
             )
             
             gr.Markdown("""
-            This is a leaderboard of one of the questions from the UGI-Leaderboard. It doesn't use the decensoring system prompt the other questions do.
+            This is a leaderboard of one of the questions from the UGI-Leaderboard. It doesn't use the decensoring system prompt the other questions do. Only the regression output is used in the UGI-Leaderboard.
             <br>
             *This leaderboard will change over time as I improve the model's predictive accuracy and as I get new data to train it on.*
             <br><br>
@@ -166,9 +166,9 @@ with GraInter:
             <br>
             **MyScore:** After generating the story, I give it a rating from 0 to 1 on how well written it was and how well it followed the prompt.
             <br>
-            **RegV1:** Using 13 unique lexical analysis metrics as the input and my scores as the output, I trained a regression model to recognize what types of writing styles people like.
+            **Reg:** Using 13 unique lexical analysis metrics as the input and my scores as the output, I trained a regression model to recognize what types of writing styles people like. To have a better seperation between model sizes, I weighted it by UGI intelligence-focused questions.
             <br>
-            **RegV2:** Basically RegV1 but slighly more style focused as oppose to RegV1 which is weighted by UGI intelligence-focused questions in order to have a better divide between model sizes.
+            **Reg+MyScore:** The average between Reg and MyScore.
             <br><br>
             Below are three of the metrics used which may be useful by themselves at detecting certain writing styles.
             <br>
@@ -227,7 +227,7 @@ with GraInter:
     def update_all_tables(query, param_ranges):
         ugi_table = update_table(leaderboard_df, query, param_ranges, UGI_COLS)
         
-        ws_df = leaderboard_df.sort_values(by='RegV1 🏆', ascending=False)
+        ws_df = leaderboard_df.sort_values(by='Reg+MyScore 🏆', ascending=False)
         ws_table = update_table(ws_df, query, param_ranges, WRITING_STYLE_COLS)
         
         arp_df = leaderboard_df.sort_values(by='Score 🏆', ascending=False)