Safety Evaluation on HarmBench 159 standard behaviors (test)
0Attack Success Rate (ASR)No attack
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| No attackModel=LLaDA-8B-Instruct2026.03 | 0 | 0 | 1 | |
| TrajHijack (core: re-mask + prefix)Model=LLaDA-8B-Instruct, Prefix length (Lg)=1282026.03 | 76.1 | 88.1 | 4.1 | |
| TrajHijack (core: re-mask + prefix)Model=Dream-7B-Instruct, Prefix length (Lg)=1282026.03 | 81.8 | 93.1 | 4.4 |