Toxicity mitigation on ToxicRandom
45.1RelevanceDATG-P
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DATG-PBase LLM=LLaMA-2 13B2024.02 | 45.1 | 43.02 | 13.4 | |
| DATG-PBase LLM=OPT 6.7B2024.02 | 44.7 | 34.25 | 16.9 | |
| DATG-PBase LLM=Alpaca 7B2024.02 | 44.2 | 57.417 | 13.5 | |
| DATG-PBase LLM=Falcon 7B2024.02 | 44.2 | 32.992 | 16.1 | |
| CONTINUATIONBase LLM=LLaMA-2 13B2024.02 | 43.9 | 32.91 | 13.1 | |
| CONTINUATIONBase LLM=OPT 6.7B2024.02 | 43.7 | 23.568 | 14.4 | |
| INJECTIONBase LLM=LLaMA-2 13B2024.02 | 43.5 | 46.191 | 14.5 | |
| CONTINUATIONBase LLM=Alpaca 7B2024.02 | 43.2 | 32.698 | 12.6 | |
| INJECTIONBase LLM=Alpaca 7B2024.02 | 43.1 | 36.36 | 14 | |
| CONTINUATIONBase LLM=Falcon 7B2024.02 | 42.9 | 25.581 | 13.7 | |
| INJECTIONBase LLM=OPT 6.7B2024.02 | 42.9 | 22.028 | 16.3 | |
| FUDGEBase LLM=Alpaca 7B2024.02 | 42.7 | 61.661 | 12.1 | |
| INJECTIONBase LLM=Falcon 7B2024.02 | 42.7 | 24.791 | 16.3 | |
| INJECTIONBase LLM=Phi-2 2.7B2024.02 | 42.7 | 23.459 | 15.4 | |
| DATG-LBase LLM=Falcon 7B2024.02 | 42.5 | 28.027 | 11.6 | |
| FUDGEBase LLM=LLaMA-2 13B2024.02 | 42.3 | 58.429 | 11.8 | |
| DATG-LBase LLM=LLaMA-2 13B2024.02 | 42.3 | 41.948 | 11.3 | |
| CONTINUATIONBase LLM=Phi-2 2.7B2024.02 | 42.3 | 21.311 | 11.2 | |
| DATG-PBase LLM=Phi-2 2.7B2024.02 | 42.2 | 38.72 | 13.4 | |
| FUDGEBase LLM=OPT 6.7B2024.02 | 42.1 | 56.963 | 14.5 | |
| FUDGEBase LLM=Falcon 7B2024.02 | 41.9 | 46.523 | 13.4 | |
| DATG-LBase LLM=Alpaca 7B2024.02 | 41.7 | 39.61 | 12 | |
| DATG-LBase LLM=OPT 6.7B2024.02 | 41.7 | 25.003 | 12.4 | |
| PREADDBase LLM=LLaMA-2 13B2024.02 | 41.5 | 61.478 | 10.7 | |
| PREADDBase LLM=OPT 6.7B2024.02 | 41.1 | 41.807 | 14.5 | |
| PREADDBase LLM=Falcon 7B2024.02 | 41 | 46.769 | 12.3 | |
| PREADDBase LLM=Alpaca 7B2024.02 | 40.9 | 55.89 | 10.7 | |
| FUDGEBase LLM=Phi-2 2.7B2024.02 | 40.7 | 42.85 | 9.6 | |
| DATG-LBase LLM=Phi-2 2.7B2024.02 | 40 | 23.119 | 9.5 | |
| PREADDBase LLM=Phi-2 2.7B2024.02 | 38.6 | 31.007 | 8.9 |