Safety Evaluation on ANTHROPIC HH (test)
88.85Safety ScoreTOSS-Pro
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TOSS-ProModel=Llama-3-8B-Instruct2026.03 | 88.85 | — | |
| TOSSModel=Llama-3-8B-Instruct2026.03 | 88.82 | — | |
| TOSS-ProModel=Llama-2-7B-Chat-hf2026.03 | 86.98 | — | |
| TOSSModel=Llama-2-7B-Chat-hf2026.03 | 83.22 | — | |
| TOSSModel=Llama-3.2-3B-Instruct2026.03 | 78.54 | — | |
| DSIRModel=Llama-3-8B-Instruct2026.03 | 67.38 | — | |
| TOSSModel=Llama-3.2-1B-Instruct2026.03 | 66.49 | — | |
| DSIRModel=Llama-3.2-1B-Instruct2026.03 | 66.03 | — | |
| DSIRModel=Llama-3.2-3B-Instruct2026.03 | 64.17 | — | |
| DSIRModel=Llama-2-7B-Chat-hf2026.03 | 63.7 | — | |
| SEALModel=Llama-3.2-3B-Instruct2026.03 | 58.94 | — | |
| SEALModel=Llama-2-7B-Chat-hf2026.03 | 58.57 | — | |
| SEALModel=Llama-3-8B-Instruct2026.03 | 58.2 | — | |
| SEALModel=Llama-3.2-1B-Instruct2026.03 | 56.89 | — | |
| SafeInstrModel=Llama-3.2-1B-Instruct2026.03 | 55.29 | — | |
| RandomModel=Llama-3-8B-Instruct2026.03 | 52.5 | — | |
| SafeInstrModel=Llama-3-8B-Instruct2026.03 | 51.49 | — | |
| SafeInstrModel=Llama-3.2-3B-Instruct2026.03 | 51.15 | — | |
| Standard SFTModel=Llama-3-8B-Instruct2026.03 | 50 | — | |
| Standard SFTModel=Llama-2-7B-Chat-hf2026.03 | 50 | — | |
| Standard SFTModel=Llama-3.2-1B-Instruct2026.03 | 50 | — | |
| Standard SFTModel=Llama-3.2-3B-Instruct2026.03 | 50 | — | |
| SafeInstrModel=Llama-2-7B-Chat-hf2026.03 | 48.16 | — | |
| RandomModel=Llama-2-7B-Chat-hf2026.03 | 48 | — | |
| DSIRModel=Llama-3-8B-Instruct2026.03 | — | 3.28 | |
| DSIRModel=Llama-2-7B-Chat-hf2026.03 | — | 3.04 | |
| RandomModel=Llama-3-8B-Instruct2026.03 | — | 2.99 | |
| RandomModel=Llama-2-7B-Chat-hf2026.03 | — | 3.08 | |
| SafeInstrModel=Llama-3-8B-Instruct2026.03 | — | 3.03 | |
| SafeInstrModel=Llama-2-7B-Chat-hf2026.03 | — | 2.93 | |
| SEALModel=Llama-3-8B-Instruct2026.03 | — | 2.94 | |
| SEALModel=Llama-2-7B-Chat-hf2026.03 | — | 2.83 | |
| Standard SFTModel=Llama-3-8B-Instruct2026.03 | — | 3.02 | |
| Standard SFTModel=Llama-2-7B-Chat-hf2026.03 | — | 2.99 | |
| TOSSModel=Llama-3-8B-Instruct2026.03 | — | 2.61 | |
| TOSSModel=Llama-2-7B-Chat-hf2026.03 | — | 2.49 | |
| TOSS-ProModel=Llama-3-8B-Instruct2026.03 | — | 2.5 | |
| TOSS-ProModel=Llama-2-7B-Chat-hf2026.03 | — | 2.14 |