Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Detoxification on AdvBench harmful behavior set
Loading...
99.17
Safety Score
PaCE
8.222
31.8335
55.445
79.0565
Jun 6, 2024
Safety Score
Updated 3mo ago
Evaluation Results
Method
Method
Links
Safety Score
PaCE
Backbone=LlaMA2-13B-Chat
2024.06
99.17
OrthoProj
Backbone=LlaMA2-13B-Chat
2024.06
96.72
PaCE
Backbone=LlaMA2-7B-Chat
2024.06
96.65
VecAdd
Backbone=LlaMA2-13B-Chat
2024.06
95.33
VecAdd
Backbone=LlaMA2-7B-Chat
2024.06
94.51
PE
Backbone=LlaMA2-13B-Chat
2024.06
93.86
OrthoProj
Backbone=LlaMA2-7B-Chat
2024.06
92.81
PE
Backbone=LlaMA2-7B-Chat
2024.06
91.9
Vanilla
Backbone=LlaMA2-13B-Chat
2024.06
18.04
Vanilla
Backbone=LlaMA2-7B-Chat
2024.06
11.72
Feedback
Search any
task
Search any
task
