Alignment Audit on Jigsaw Toxic Comment
0Average Treatment Effect (ATE)SAE-based explanation
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAE-based explanationtop_k_features=12025.06 | 0 | — | |
| SAE-based explanationtop_k_features=22025.06 | 0 | — | |
| SAE-based explanationtop_k_features=32025.06 | 0.02 | — | |
| SAE-based explanationtop_k_features=42025.06 | 0.03 | — | |
| SAE-based explanationtop_k_features=52025.06 | 0.04 | — |