Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Misalignment Evaluation on CAFT 800 general prompts
Loading...
0
Misalignment Rate
GRASP (2 epochs)
-0.16
0.92
2
3.08
May 26, 2026
Misalignment Rate
Betley Misalignment Rate
Updated 6d ago
Evaluation Results
Method
Method
Links
Misalignment Rate
Betley Misalignment Rate
GRASP (2 epochs)
Probe=unsupervised, In...
2026.05
0
-
GRASP (3 epochs)
Probe=unsupervised, In...
2026.05
0
-
Inference-ablate vsvd
Inference-time ablatio...
2026.05
0.3
0.3
CAFT-PCA
Probe=labelled, Interv...
2026.05
0.5
-
GRASP
Epochs=1 epoch
2026.05
0.5
0.6
v-ref (1 epoch)
Probe=labelled, Interv...
2026.05
0.6
-
CAFT-time vsvd
Epochs=1 epoch
2026.05
1.3
1.1
CAFT-SAE
Probe=labelled, Interv...
2026.05
2.4
-
Naive
Epochs=1 epoch
2026.05
2.7
2.5
v-ref (2 epochs)
Probe=labelled, Interv...
2026.05
3.3
-
Naive (corruption baseline)
Probe=—, Intervention=—
2026.05
4
-
Feedback
Search any
task
Search any
task