DontPlanToEnd
commited on
Commit
β’
481e529
1
Parent(s):
fb27588
Update app.py
Browse files
app.py
CHANGED
@@ -60,16 +60,22 @@ with demo:
|
|
60 |
gr.Markdown("## UGI Leaderboard", elem_classes="text-lg text-center")
|
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 |
Willingness: A more narrow score, solely measuring the LLM's willingness to answer controversial questions.
|
|
|
64 |
Unruly: Knowledge of activities that are generally frowned upon.
|
|
|
65 |
Internet: Knowledge of various internet information, from professional to deviant.
|
|
|
66 |
CrimeStats: Knowledge of crime statistics which are uncomfortable to talk about.
|
|
|
67 |
Stories/Jokes: Ability to write offensive stories and jokes.
|
|
|
68 |
PolContro: Knowledge of politically/socially controversial information.
|
69 |
""")
|
70 |
with gr.Column():
|
71 |
with gr.Row():
|
72 |
-
search_bar = gr.Textbox(placeholder=" π Search for a model...", show_label=False)
|
73 |
with gr.Row():
|
74 |
filter_columns_size = gr.CheckboxGroup(
|
75 |
label="Model sizes (in billions of parameters)",
|
|
|
60 |
gr.Markdown("## UGI Leaderboard", elem_classes="text-lg text-center")
|
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 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():
|
77 |
with gr.Row():
|
78 |
+
search_bar = gr.Textbox(placeholder=" π Search for a model...", show_label=False, elem_id="search-bar")
|
79 |
with gr.Row():
|
80 |
filter_columns_size = gr.CheckboxGroup(
|
81 |
label="Model sizes (in billions of parameters)",
|