List your company

Identity Shifting Attack on HEx-PHI Identity Shifting Attack (300 questions)

22.83ASRSFT
-0.91325.250911.41517.5791Feb 19, 2026

Evaluation results

MethodLinks
SFTBackbone=Qwen2.5-7B-Instruct2026.0222.83
SFTBackbone=Qwen2.5-3B-Instruct2026.0220.67
C-SFTBackbone=Qwen2.5-7B-Instruct2026.0220.17
SFTBackbone=Meta-Llama-3.1-8B2026.0219.67
SFTBackbone=Phi-3.5-mini-instruct2026.0218
C-SFTBackbone=Qwen2.5-3B-Instruct2026.0216.5
A-Reg.Backbone=Qwen2.5-7B-Instruct2026.0216
A-Reg.Backbone=Qwen2.5-3B-Instruct2026.0215.84
C-SFTBackbone=Phi-3.5-mini-instruct2026.0212
A-Reg.Backbone=Phi-3.5-mini-instruct2026.028
SFTBackbone=Llama-3.2-3B-Instruct2026.024.33
InitialBackbone=Qwen2.5-3B-Instruct2026.024.33
A-Reg.Backbone=Llama-3.2-3B-Instruct2026.023.5
InitialBackbone=Qwen2.5-7B-Instruct2026.023.44
C-SFTBackbone=Meta-Llama-3.1-8B2026.022.67
InitialBackbone=Phi-3.5-mini-instruct2026.022.44
C-SFTBackbone=Llama-3.2-3B-Instruct2026.021
A-Reg.Backbone=Meta-Llama-3.1-8B2026.020.83
InitialBackbone=Meta-Llama-3.1-8B2026.020
InitialBackbone=Llama-3.2-3B-Instruct2026.020