List your company

Emergent Misalignment Measurement on Legal

0.58MisalignmentPersona Vectors
-1.039.837520.70531.5725Aug 8, 2025

Evaluation results

MethodLinks
Persona VectorsModel=Qwen2.5-32B2025.080.580.25
InterleavingModel=Qwen2.5-32B2025.083.217.59
Interleaving++Model=Qwen2.5-32B2025.084.016.64
KLModel=Qwen2.5-32B2025.085.830
Interleaving+Model=Qwen2.5-32B2025.0813.7713.03
MisalignedModel=Qwen2.5-32B2025.0840.839.17