Activation Steering on DEBATEMIX
56Avg. Activation ChangeK-Steer
Evaluation Results
| Method | Links | |
|---|---|---|
| K-SteerK=1, Model=Olmo2025.05 | 56 | |
| K-SteerK=1, Model=Llama2025.05 | 56 | |
| CAAK=1, Model=Olmo2025.05 | 43 | |
| DCTK=1, Model=Olmo2025.05 | 32 | |
| DCTK=1, Model=Llama2025.05 | 29 | |
| K-SteerK=2, Model=Olmo2025.05 | 25 | |
| K-SteerK=2, Model=Llama2025.05 | 25 | |
| CAAK=2, Model=Olmo2025.05 | 24 | |
| DCTK=2, Model=Llama2025.05 | 20 | |
| DCTK=2, Model=Olmo2025.05 | 16 | |
| K-SteerK=3, Model=Olmo2025.05 | 9 | |
| K-SteerK=3, Model=Llama2025.05 | 9 | |
| CAAK=3, Model=Llama2025.05 | 8 | |
| CAAK=2, Model=Llama2025.05 | 7 | |
| DCTK=3, Model=Llama2025.05 | 7 | |
| CAAK=3, Model=Olmo2025.05 | 4 | |
| CAAK=1, Model=Llama2025.05 | 2 | |
| DCTK=3, Model=Olmo2025.05 | 1 |