Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
3-class classification on Narratives 54 narratives, 48,640 samples (held-out)
Loading...
57.8
Aligned Accuracy
SVAF (neural)
40.536
45.018
49.5
53.982
Apr 5, 2026
Aligned Accuracy
Guarded Accuracy
Rejected Accuracy
Overall Accuracy
Updated 1mo ago
Evaluation Results
Method
Method
Links
Aligned Accuracy
Guarded Accuracy
Rejected Accuracy
Overall Accuracy
SVAF (neural)
learned_encoder=true,...
2026.04
57.8
87.7
70.5
78.7
Heuristic per-field (αf)
fusion_gate=weighted c...
2026.04
48.9
83.4
63.2
73.1
Scalar + temporal decay
temporal_decay=true
2026.04
43.5
80.1
55.8
68.4
Scalar
type=cosine, theta (θ)...
2026.04
41.2
79.3
52.1
66.8
Feedback
Search any
task
Search any
task