Safety Evaluation on SLIMORCA OOD Safety Scenario (test)
2.4Harmfulness ScoreTOSS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TOSSBackbone=Llama-3-8B-Instruct, OOD setting=true2026.03 | 2.4 | 53.03 | 66.59 | |
| Standard SFTBackbone=Llama-3-8B-Instruct, OOD setting=true2026.03 | 4.17 | 87.57 | 50 |