DontPlanToEnd commited on
Commit
915b360
1 Parent(s): 0e29ede

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +16 -17
app.py CHANGED
@@ -54,17 +54,23 @@ def update_table(df: pd.DataFrame, query: str, param_ranges: dict) -> pd.DataFra
54
  return filtered_df[UGI_COLS] # Return only the columns defined in UGI_COLS
55
 
56
  # Define the Gradio interface
57
- GraInterface = gr.Blocks()
58
 
59
- with GraInterface:
60
- gr.Markdown("## UGI Leaderboard", elem_classes="text-lg text-center", elem_id="title")
61
  gr.Markdown("""
62
- UGI: Uncensored General Intelligence. The average of 5 different subjects that LLMs are commonly steered away from. The leaderboard is made from roughly 60 questions overall, measuring both "willingness to answer" and "accuracy" in fact-based controversial questions.\n
63
- Willingness: A more narrow, 10-point score, solely measuring the LLM's willingness to answer controversial questions.\n
64
- Unruly: Knowledge of activities that are generally frowned upon.\n
65
- Internet: Knowledge of various internet information, from professional to deviant.\n
66
- CrimeStats: Knowledge of crime statistics which are uncomfortable to talk about.\n
67
- Stories/Jokes: Ability to write offensive stories and jokes.\n
 
 
 
 
 
 
68
  PolContro: Knowledge of politically/socially controversial information.
69
  """)
70
  with gr.Column():
@@ -112,13 +118,6 @@ with GraInterface:
112
  inputs=inputs,
113
  outputs=outputs
114
  )
115
-
116
- # Add custom CSS styles
117
- GraInterface.css = """
118
- #title {
119
- text-align: center;
120
- }
121
- """
122
 
123
  # Launch the Gradio app
124
- GraInterface.launch()
 
54
  return filtered_df[UGI_COLS] # Return only the columns defined in UGI_COLS
55
 
56
  # Define the Gradio interface
57
+ demo = gr.Blocks()
58
 
59
+ with demo:
60
+ gr.HTML("""<h1 align="center">UGI Leaderboard</h1>""")
61
  gr.Markdown("""
62
+ UGI: Uncensored General Intelligence. The average of 5 different subjects that LLMs are commonly steered away from. The leaderboard is made from roughly 60 questions overall, measuring both "willingness to answer" and "accuracy" in fact-based controversial questions.
63
+
64
+ Willingness: A more narrow, 10-point score, solely measuring the LLM's willingness to answer controversial questions.
65
+
66
+ Unruly: Knowledge of activities that are generally frowned upon.
67
+
68
+ Internet: Knowledge of various internet information, from professional to deviant.
69
+
70
+ CrimeStats: Knowledge of crime statistics which are uncomfortable to talk about.
71
+
72
+ Stories/Jokes: Ability to write offensive stories and jokes.
73
+
74
  PolContro: Knowledge of politically/socially controversial information.
75
  """)
76
  with gr.Column():
 
118
  inputs=inputs,
119
  outputs=outputs
120
  )
 
 
 
 
 
 
 
121
 
122
  # Launch the Gradio app
123
+ demo.launch()