Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Concept-based Steering on AXBENCH (test)

1.102Overall Steering Score

PrOSV

-0.019120.271940.5630.85406May 7, 2026
Updated 26d ago

Evaluation Results

MethodLinks
2026.05
1.102
2026.05
1.075
2026.05
1.06
2026.05
1.049
2026.05
0.982
2026.05
0.947
2026.05
0.919
2026.05
0.905
2026.05
0.892
2026.05
0.886
2026.05
0.859
2026.05
0.821
2026.05
0.803
2026.05
0.798
2026.05
0.788
2026.05
0.777
2026.05
0.769
2026.05
0.758
2026.05
0.756
2026.05
0.736
2026.05
0.701
2026.05
0.698
2026.05
0.663
2026.05
0.546
2026.05
0.322
2026.05
0.297
2026.05
0.072
2026.05
0.024