Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Alignment Evaluation on WildJailbreak
Loading...
98.6
Safety Rate
Stair-DPO
41.192
56.096
71
85.904
May 30, 2026
Safety Rate
Updated 1d ago
Evaluation Results
Method
Method
Links
Safety Rate
Stair-DPO
Architecture=Qwen3-30B...
2026.05
98.6
MESA
Architecture=Qwen3-30B...
2026.05
97.65
SafeX
Architecture=Qwen3-30B...
2026.05
96.35
SFT
Architecture=Qwen3-30B...
2026.05
95
GRPO
Architecture=Qwen3-30B...
2026.05
94.4
MESA
Architecture=DeepSeek-...
2026.05
90.9
Stair-DPO
Architecture=DeepSeek-...
2026.05
83.6
Stair-SFT
Architecture=DeepSeek-...
2026.05
77.9
SFT
Architecture=DeepSeek-...
2026.05
77.7
Base(instruct)
Architecture=Qwen3-30B...
2026.05
75.3
SafeX
Architecture=DeepSeek-...
2026.05
64
Stair-SFT
Architecture=Qwen3-30B...
2026.05
62.65
GRPO
Architecture=DeepSeek-...
2026.05
44.1
Base(chat)
Architecture=DeepSeek-...
2026.05
43.4
Feedback
Search any
task
Search any
task