Steerability Correlation Analysis on Binary Concept Families Controlled, Alin > 0.1
0.93Correlation Coefficient (rho)Qwen-1.5B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen-1.5B2026.04 | 0.93 | 10 | |
| Qwen-7B2026.04 | 0.86 | 13 | |
| Llama-8B2026.04 | 0.84 | 14 | |
| Gemma-2-2B2026.04 | 0.52 | 12 | |
| Pythia-2.8B2026.04 | 0.5 | 9 |