Backdoor Attack on AdvBench
91.8ASRBadDLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BadDLMModel=LLaDA-8B-Instruct2026.05 | 91.8 | 65.3 | |
| RL-basedModel=LLaDA-8B-Instruct2026.05 | 65.2 | 64.3 | |
| SFT-basedModel=LLaDA-8B-Instruct2026.05 | 48.4 | 65.4 | |
| Benign (No Attack)Model=LLaDA-8B-Instruct2026.05 | 1.2 | 65.5 |