Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Political Bias Assessment on 600 non-financial political prompts
Loading...
47.3
P(R)
Naive (corruption baseline)
33.052
36.751
40.45
44.149
May 26, 2026
P(R)
Delta P(R)
Updated 6d ago
Evaluation Results
Method
Method
Links
P(R)
Delta P(R)
Naive (corruption baseline)
Probe=–, Model size=1....
2026.05
47.3
0.137
CAFT-time vsvd (top-10 AUROC)
Probe=unsupervised, Mo...
2026.05
45.9
0.123
Inference-ablate vsvd (top-10 AUROC)
Probe=unsupervised, Mo...
2026.05
41.6
0.08
GRASP
Probe=unsupervised, Mo...
2026.05
39.7
0.061
Pretrained
Probe=–, Model size=1....
2026.05
33.6
0
Feedback
Search any
task
Search any
task