Safety Evaluation on HHH
63.9HHH ScoreFL + Safety filter + CAI
Evaluation Results
| Method | Links | |
|---|---|---|
| FL + Safety filter + CAIFL method=SCAFFOLD2025.02 | 63.9 | |
| FL + Safety filter + CAIFL method=FedAvg2025.02 | 63.7 | |
| FL + CAIFL method=SCAFFOLD2025.02 | 62.6 | |
| Llama3.1-8B-InstructFL method=FedAvg2025.02 | 60 | |
| Llama3.1-8B-InstructFL method=SCAFFOLD2025.02 | 60 | |
| FL + CAIFL method=FedAvg2025.02 | 57.3 | |
| FL + Safety filterFL method=SCAFFOLD2025.02 | 54.6 | |
| FL + Safety filterFL method=FedAvg2025.02 | 51.8 | |
| FLFL method=SCAFFOLD2025.02 | 49.5 | |
| FLFL method=FedAvg2025.02 | 49.3 |