List your company
ResearchFormerly WizWand

Jailbreak Robustness on Fortress

7.2ASRPolicyAlign
5.04819.57434.148.626Jun 24, 2026
Updated 4mo ago

Evaluation results

MethodLinks
PolicyAlignBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=Policy-Based Safety Alignment2026.067.2
PolicyAlignBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=Policy-Based Safety Alignment2026.067.4
SFTBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=Supervised Fine-Tuning2026.067.6
GRPO+PolicyBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=Group Relative Policy Optimization with policy-conditioned reward2026.068.6
GRPO+PolicyBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=Group Relative Policy Optimization with policy-conditioned reward2026.069.2
ICLBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=In-Context Learning (System Prompt)2026.069.4
AlphaAlignBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=RL-based alignment2026.069.4
NSPOBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=Null-space constrained policy optimization2026.069.8
AlphaAlignBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=RL-based alignment2026.0610
SFTBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=Supervised Fine-Tuning2026.0610.2
NSPOBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=Null-space constrained policy optimization2026.0610.4
PolicyAlignBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=Policy-Based Safety Alignment2026.0615.4
GRPO+PolicyBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=Group Relative Policy Optimization with policy-conditioned reward2026.0616.4
ICLBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=In-Context Learning (System Prompt)2026.0616.6
SFTBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=Supervised Fine-Tuning2026.0617.8
AlphaAlignBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=RL-based alignment2026.0618.2
NSPOBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=Null-space constrained policy optimization2026.0618.4
BaseBackbone=LLaMA-3.2-3B-Instruct, Alignment Strategy=Base2026.0632.4
ICLBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=In-Context Learning (System Prompt)2026.0643.2
BaseBackbone=Qwen2.5-14B-Instruct, Alignment Strategy=Base2026.0646.4
BaseBackbone=Qwen2.5-7B-Instruct, Alignment Strategy=Base2026.0661