Safety Alignment on WildJailbreak (Safe@1)
77.2Safe@1R1 - 8B + UnsafeChain full
Evaluation Results
| Method | Links | |
|---|---|---|
| R1 - 8B + UnsafeChain fullSize=8B, Alignment=UnsafeChain (full)2025.07 | 77.2 | |
| R1 - 7B + UnsafeChain fullSize=7B, Alignment=UnsafeChain (full)2025.07 | 74.8 | |
| R1 - 8B + UnsafeChain selectedSize=8B, Alignment=UnsafeChain (selected subset)2025.07 | 73.6 | |
| R1 - 8B + SafeChainSize=8B, Alignment=SafeChain2025.07 | 68 | |
| R1 - 8B + STAR-1Size=8B, Alignment=STAR-12025.07 | 67.6 | |
| R1 - 8B + UnsafeChain randomSize=8B, Alignment=UnsafeChain (random subset)2025.07 | 65.6 | |
| R1 - 7B + STAR-1Size=7B, Alignment=STAR-12025.07 | 65.2 | |
| R1 - 8BSize=8B2025.07 | 64.8 | |
| R1 - 1.5B + UnsafeChain fullSize=1.5B, Alignment=UnsafeChain (full)2025.07 | 64.8 | |
| R1 - 7B + SafeChainSize=7B, Alignment=SafeChain2025.07 | 62.4 | |
| R1 - 7B + UnsafeChain selectedSize=7B, Alignment=UnsafeChain (selected subset)2025.07 | 62 | |
| R1 - 7B + UnsafeChain randomSize=7B, Alignment=UnsafeChain (random subset)2025.07 | 61.6 | |
| R1 - 7BSize=7B2025.07 | 60.8 | |
| SIFTRuntime (min)=215.62026.04 | 56 | |
| BLURRuntime (min)=114.72026.04 | 54.4 | |
| AdamWRuntime (min)=110.32026.04 | 52.8 | |
| MuonRuntime (min)=191.22026.04 | 52 | |
| POMERuntime (min)=115.42026.04 | 51.6 | |
| R1 - 1.5B + SafeChainSize=1.5B, Alignment=SafeChain2025.07 | 50.8 | |
| Base modelRuntime (min)=N/A2026.04 | 47.6 | |
| R1 - 1.5BSize=1.5B2025.07 | 38.4 | |
| R1 - 1.5B + UnsafeChain randomSize=1.5B, Alignment=UnsafeChain (random subset)2025.07 | 37.2 | |
| R1 - 1.5B + STAR-1Size=1.5B, Alignment=STAR-12025.07 | 36.4 | |
| R1 - 1.5B + UnsafeChain selectedSize=1.5B, Alignment=UnsafeChain (selected subset)2025.07 | 35.2 |