Llama-Guard
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Llama-Guard 3-8B
2.38ASR
56
Llama-Guard
82.14Harmfulness (%)
36
LLaMA Guard 8B 3.1
2.16Mean PPL
10
LLaMA Guard 2 8B
2.34PPL Mean
10
LLaMA Guard 7B
3.05Mean Perplexity (PPL)
10
LLaMA Guard 2 8B
100Bypass Rate
1
LLaMA Guard 7B
98.65Bypass Rate
1
Llama Guard LLM latent mechanism
—Primary metric
0