Emergent Misalignment Measurement on Medical General Evaluation
0.38MisalignmentPersona Vectors
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Persona VectorsModel=Qwen2.5-32B2025.08 | 0.38 | 0 | |
| KLModel=Qwen2.5-32B2025.08 | 2.5 | 0 | |
| Interleaving++Model=Qwen2.5-32B2025.08 | 7.89 | 7.94 | |
| InterleavingModel=Qwen2.5-32B2025.08 | 11.33 | 7.42 | |
| Interleaving+Model=Qwen2.5-32B2025.08 | 16.74 | 9.21 | |
| MisalignedModel=Qwen2.5-32B2025.08 | 35.42 | 5.42 |