Spaces:
Running
Running
Joshua Saxe
commited on
Commit
•
164ca0b
1
Parent(s):
2e9a66b
adding descriptive text
Browse files
app.py
CHANGED
@@ -92,7 +92,7 @@ def display_results(df, title, description,include_barchart=True):
|
|
92 |
# Continue displaying the results for each test
|
93 |
display_results(interpreter_df, 'LLMs Assistance to Hack Code Interpreter Environments', 'The plot below shows how often each model complies with attempts to hack a connected code interpreter. Lower values indicate safer models.', False)
|
94 |
display_results(prompt_df, 'LLMs Susceptibility to Prompt Injection Attacks', 'The plot below shows how often each model succumbs to prompt injection attacks, categorized into threat categories detailed in our paper. Lower values indicate safer models.', False)
|
95 |
-
display_results(exploit_df, 'LLMs Capability to Solve Cyber Capture the Flag Challenges', 'The plot below shows how well each model performs at solving capture-the-flag style program exploitation challenges. Higher values indicate more capable models.',
|
96 |
|
97 |
# Load additional JSON files
|
98 |
data_mitre = json.load(open("mitre.json"))
|
|
|
92 |
# Continue displaying the results for each test
|
93 |
display_results(interpreter_df, 'LLMs Assistance to Hack Code Interpreter Environments', 'The plot below shows how often each model complies with attempts to hack a connected code interpreter. Lower values indicate safer models.', False)
|
94 |
display_results(prompt_df, 'LLMs Susceptibility to Prompt Injection Attacks', 'The plot below shows how often each model succumbs to prompt injection attacks, categorized into threat categories detailed in our paper. Lower values indicate safer models.', False)
|
95 |
+
display_results(exploit_df, 'LLMs Capability to Solve Cyber Capture the Flag Challenges', 'The plot below shows how well each model performs at solving capture-the-flag style program exploitation challenges. Higher values indicate more capable models.', False)
|
96 |
|
97 |
# Load additional JSON files
|
98 |
data_mitre = json.load(open("mitre.json"))
|