Safety on WildJailbreak
17.6Harmful Response RatioSTAR-1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| STAR-1Model Size=8B2026.01 | 17.6 | — | |
| THINKSAFEModel Size=8B2026.01 | 21.15 | — | |
| SafeKeyModel Size=8B2026.01 | 30.85 | — | |
| DirectRefusalModel Size=8B2026.01 | 32.6 | — | |
| THINKSAFEModel Size=7B2026.01 | 35.4 | — | |
| THINKSAFEModel Size=1.5B2026.01 | 40.5 | — | |
| SafeChainModel Size=8B2026.01 | 42.45 | — | |
| SafeKeyModel Size=7B2026.01 | 45.75 | — | |
| SafeChainModel Size=7B2026.01 | 45.85 | — | |
| SafeChainModel Size=1.5B2026.01 | 46.95 | — | |
| SafeKeyModel Size=1.5B2026.01 | 47.65 | — | |
| STAR-1Model Size=7B2026.01 | 48.75 | — | |
| STAR-1Model Size=1.5B2026.01 | 49.65 | — | |
| InitialModel Size=8B2026.01 | 49.7 | — | |
| DirectRefusalModel Size=7B2026.01 | 50.2 | — | |
| SafePathModel Size=8B2026.01 | 50.2 | — | |
| DirectRefusalModel Size=1.5B2026.01 | 50.45 | — | |
| InitialModel Size=1.5B2026.01 | 51.55 | — | |
| SafePathModel Size=1.5B2026.01 | 51.8 | — | |
| SafePathModel Size=7B2026.01 | 52.65 | — | |
| InitialModel Size=7B2026.01 | 53.15 | — | |
| DeepSeek-V3.2Mode=Non-Reasoning, Architecture=MoE, Total Params=671B, Activated Params=37B2026.01 | — | 76.8 | |
| DGR SFTAlignment Dataset=DirectRefusal, Backbone=s1.1-7B2026.02 | — | 91.2 | |
| DGR SFTAlignment Dataset=STAR-1, Backbone=s1.1-7B2026.02 | — | 97.2 | |
| DGR SFTAlignment Dataset=R1-ACT, Backbone=s1.1-7B2026.02 | — | 100 | |
| EXAONE 4.0Mode=Non-Reasoning, Architecture=Dense, Total Params=32B, Activated Params=32B2026.01 | — | 49.6 | |
| K-EXAONEMode=Non-Reasoning, Architecture=MoE, Total Params=236B, Activated Params=23B2026.01 | — | 91.6 | |
| Qwen2.5-7B-Instruct2026.02 | — | 56.8 | |
| Qwen3-235B-A22B Instruct-2507Architecture=MoE, Total Params=235B, Activated Params=22B2026.01 | — | 91.1 | |
| s1.1-7B2026.02 | — | 48.8 | |
| Vanilla SFTAlignment Dataset=DirectRefusal, Backbone=s1.1-7B2026.02 | — | 98.8 | |
| Vanilla SFTAlignment Dataset=STAR-1, Backbone=s1.1-7B2026.02 | — | 95.6 | |
| Vanilla SFTAlignment Dataset=R1-ACT, Backbone=s1.1-7B2026.02 | — | 100 |