Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Instruction Following on HarmBench Clean
Loading...
99
Clean Rate
Base (8B Instruct)
9.56
32.78
56
79.22
May 8, 2026
Clean Rate
Updated 22d ago
Evaluation Results
Method
Method
Links
Clean Rate
Base (8B Instruct)
Backbone=Llama 3
2026.05
99
LPA (ours)
Backbone=Llama 3
2026.05
99
L3-LAT
Backbone=Llama 3
2026.05
92
L3-CAT
Backbone=Llama 3
2026.05
92
LPA-overfit (ours)
Backbone=Llama 3
2026.05
88
LPA (ours)
Backbone=Llama 2
2026.05
79
Base (7B chat-hf)
Backbone=Llama 2
2026.05
78
L2-CAT
Backbone=Llama 2
2026.05
67
LPA-overfit (ours)
Backbone=Llama 2
2026.05
60
L2-LAT
Backbone=Llama 2
2026.05
13
Feedback
Search any
task
Search any
task