List your company

Multi-Turn Jailbreak Robustness on Actor Attack/MUSE-A/TRACE Scenarios

47.17Actor Attack ASRQwen2.5-7B-IT
1.378813.266925.15537.0431May 9, 2026

Evaluation results

MethodLinks
Qwen2.5-7B-IT2026.0547.1751.5787.8458.63
MUSE-DBackbone=Qwen2.5-7B-IT2026.0530.816.2850.9429.34
TRACEBackbone=Qwen2.5-7B-IT, variant=w/o lat.2026.0517.611.2610.069.64
TRACEBackbone=Qwen2.5-7B-IT, variant=ours2026.0515.721.899.439.01
SafeMTBackbone=Qwen2.5-7B-IT2026.053.149.4328.313.62