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

Activation Steering on DMD2 (held-out concepts)

46.7Input Fidelity

Unsteered

27.77232.68637.642.514May 7, 2026
Updated 22d ago

Evaluation Results

MethodLinks
2026.05
46.720.4
2026.05
44.626.2
2026.05
43.127.8
2026.05
41.628
2026.05
35.532.2
2026.05
30.527.5
2026.05
28.527.5