Win Rate Evaluation on ANTHROPIC HH (test)
88.82Win RateTOSS
Evaluation Results
| Method | Links | |
|---|---|---|
| TOSSComparison model=Standard SFT, Backbone=Llama-3-8B-Instruct2026.03 | 88.82 | |
| TOSSParadigm=Training-based defense, Base Model=Llama-3-8B-Instruct2026.03 | 88.82 | |
| TOSSComparison model=SEAL, Backbone=Llama-3-8B-Instruct2026.03 | 84.54 | |
| SafetyLockParadigm=Training-free defense, Base Model=Llama-3-8B-Instruct2026.03 | 50.25 | |
| Standard SFTParadigm=/, Base Model=Llama-3-8B-Instruct2026.03 | 50 | |
| Circuit BreakersParadigm=Training-based defense, Base Model=Llama-3-8B-Instruct2026.03 | 40.25 |