Question Answering on TruthfulQA (ROUGE Metrics)
35ROUGE-1Dream
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DreamDefense=None2026.05 | 35 | 21 | 31 | |
| LLaDADefense=None2026.05 | 30 | 18 | 28 | |
| DiffuGuardBase Model=LLaDA2026.05 | 27 | 15 | 26 | |
| Adaptive Steering and RemaskingBase Model=LLaDA2026.05 | 21 | 14 | 19 | |
| Self-reminderBase Model=LLaDA2026.05 | 18 | 8 | 17 | |
| DiffuGuardBase Model=Dream2026.05 | 11 | 4 | 10 | |
| Self-reminderBase Model=Dream2026.05 | 11 | 5 | 10 | |
| Adaptive Steering and RemaskingBase Model=Dream2026.05 | 11 | 6 | 16 |