Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Self-attention inverse temperature scaling analysis on Proof-Pile-2 (PP2)
Loading...
7.2
Tie Percentage
Qwen7B
6.784
6.892
7
7.108
May 12, 2026
Tie Percentage
Xi*Beta (Section 3)
Xi_hat_Lambda
Xi_hat_Alpha
Xi_hat_Delta
Updated 20d ago
Evaluation Results
Method
Method
Links
Tie Percentage
Xi*Beta (Section 3)
Xi_hat_Lambda
Xi_hat_Alpha
Xi_hat_Delta
Qwen7B
neval/iter=32768
2026.05
7.2
2
1.74
1.68
2.42
Qwen7B
neval/iter=16384
2026.05
6.8
2.5
2.33
1.4
2.73
Feedback
Search any
task
Search any
task
