Open Generation Benchmark
99.33Detox ScoreSTA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| STAModel=Gemma2-9B-it, #Acts=35842026.02 | 99.33 | 25 | |
| AUSteer-FFNModel=Gemma2-9B-it, #Acts=<= 1002026.02 | 99.25 | 30 | |
| SADIModel=Gemma2-9B-it, #Acts=kh * 2562026.02 | 99.08 | 25.5 | |
| AUSteer-HeadModel=Gemma2-9B-it, #Acts=<= 1002026.02 | 99 | 26.5 | |
| ITIModel=Gemma2-9B-it, #Acts=kh * 2562026.02 | 98.83 | 23 | |
| CAAModel=Gemma2-9B-it, #Acts=35842026.02 | 98.75 | 23.5 | |
| AUSteer-HeadModel=LLaMA2-7B-Chat, #Acts=<= 1002026.02 | 89.66 | 16 | |
| AUSteer-FFNModel=LLaMA2-7B-Chat, #Acts=<= 1002026.02 | 89.24 | 22 | |
| SADIModel=LLaMA2-7B-Chat, #Acts=kh * 1282026.02 | 86.32 | 13.5 | |
| CAAModel=LLaMA2-7B-Chat, #Acts=40962026.02 | 84.57 | 11 | |
| ITIModel=LLaMA2-7B-Chat, #Acts=kh * 1282026.02 | 83.49 | 12.5 | |
| AUSteer-FFNModel=Qwen3-8B, #Acts=<= 1002026.02 | 81.65 | 34 | |
| AUSteer-HeadModel=Qwen3-8B, #Acts=<= 1002026.02 | 80.9 | 26 | |
| SADIModel=Qwen3-8B, #Acts=kh * 1282026.02 | 78.07 | 25.5 | |
| CAAModel=Qwen3-8B, #Acts=40962026.02 | 76.23 | 27 | |
| ITIModel=Qwen3-8B, #Acts=kh * 1282026.02 | 75.15 | 26 |