Activation Steering on AxBench Gemma-2-2B layer 20
0.871Steering ScoreA-PSRMSE
Evaluation Results
| Method | Links | |
|---|---|---|
| A-PSRMSEMethod Type=Multi-rank and/or multi-layer2026.05 | 0.871 | |
| LoReFT-RePSMethod Type=Multi-rank and/or multi-layer2026.05 | 0.805 | |
| LoRA-RePSMethod Type=Multi-rank and/or multi-layer2026.05 | 0.793 | |
| A-ConstLLMethod Type=Multi-rank and/or multi-layer2026.05 | 0.792 | |
| A-ConstMSEMethod Type=Multi-rank and/or multi-layer2026.05 | 0.783 | |
| Hyper-SteerMethod Type=Multi-rank and/or multi-layer2026.05 | 0.742 | |
| promptMethod Type=Multi-rank and/or multi-layer2026.05 | 0.731 | |
| SFTMethod Type=Multi-rank and/or multi-layer2026.05 | 0.714 | |
| A-PSRLLMethod Type=Multi-rank and/or multi-layer2026.05 | 0.69 | |
| LoRAMethod Type=Multi-rank and/or multi-layer2026.05 | 0.641 | |
| S-PSRLLRank=1, Intervention type=single-layer2026.05 | 0.618 | |
| ΦSV,r=1Rank=1, Intervention type=single-layer2026.05 | 0.606 | |
| ReFT-r1Rank=1, Intervention type=single-layer2026.05 | 0.509 | |
| S-ConstLLRank=1, Intervention type=single-layer2026.05 | 0.504 | |
| S-PSRMSERank=1, Intervention type=single-layer2026.05 | 0.367 | |
| S-ConstMSERank=1, Intervention type=single-layer2026.05 | 0.311 | |
| S-ConstDiMRank=1, Intervention type=single-layer2026.05 | 0.178 | |
| SAERank=1, Intervention type=single-layer2026.05 | 0.151 |