Utility Evaluation on XSTest Safe Prompts
97.2ComplianceFedDPO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FedDPOSetting=Poisoned2026.04 | 97.2 | 2.8 | |
| FedDetoxSetting=Ours2026.04 | 94.4 | 5.6 | |
| Qwen2.5-1.5B-InstructSetting=Original2026.04 | 92.8 | 7.2 |
| Method | Links | ||
|---|---|---|---|
| FedDPOSetting=Poisoned2026.04 | 97.2 | 2.8 | |
| FedDetoxSetting=Ours2026.04 | 94.4 | 5.6 | |
| Qwen2.5-1.5B-InstructSetting=Original2026.04 | 92.8 | 7.2 |