Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Refusal Induction on Refusal Induction held-in and held-out
Loading...
55
Activation Patching
Difference-in-means
8.2
20.35
32.5
44.65
Feb 17, 2026
Activation Patching
Attribution Patching
Attention Head Knockouts
ITI Score
Random Selections Score
Updated 4d ago
Evaluation Results
Method
Method
Links
Activation Patching
Attribution Patching
Attention Head Knockouts
ITI Score
Random Selections Score
Difference-in-means
Model=OLMo-13B, alpha=...
2026.02
55
55
41
58
28
Mean steering
Model=Qwen-14B, alpha=...
2026.02
48
45
8
18
20
Mean steering
Model=OLMo-13B, alpha=...
2026.02
48
49
22
34
30
Difference-in-means
Model=Qwen-14B, alpha=...
2026.02
46
41
27
20
25
ReFT
Model=Qwen-14B, alpha=...
2026.02
40
34
6
23
5
ReFT
Model=OLMo-13B, alpha=...
2026.02
24
24
14
31
27
Difference-in-means
Model=SOLAR-10.7B, alp...
2026.02
24
23
20
14
13
Mean steering
Model=SOLAR-10.7B, alp...
2026.02
12
10
12
6
6
ReFT
Model=SOLAR-10.7B, alp...
2026.02
10
8
10
5
5
Feedback
Search any
task
Search any
task