Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Majority Vote on Majority-vote 432 prompts (test)
Loading...
100
Accuracy
Soft DPS
47.376
61.038
74.7
88.362
May 5, 2026
Accuracy
Absolute Cosine Similarity (|cos|)
R2 Score
Updated 28d ago
Evaluation Results
Method
Method
Links
Accuracy
Absolute Cosine Similarity (|cos|)
R2 Score
Soft DPS
logit offset α=5
2026.05
100
-
-
Soft DPS
logit offset α=10
2026.05
100
-
-
Soft DPS
logit offset α=50
2026.05
100
-
-
Baseline (vanilla Qwen3-8B)
Condition=vanilla Qwen...
2026.05
51.4
-
-
Random direction
logit offset α=10, see...
2026.05
49.4
-
-
Best probe (layer 2)
layer=2
2026.05
-
0.003
0.9998
Feedback
Search any
task
Search any
task