Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Question Answering on MMLU-Pro n=8312 (test) (Act, Acc, WA)
Loading...
97.6
Activation (Act)
CRC
26.568
45.009
63.45
81.891
Jun 28, 2026
Activation (Act)
Accuracy (Acc)
Weighted Average (WA)
Updated 26d ago
Evaluation Results
Method
Method
Links
Activation (Act)
Accuracy (Acc)
Weighted Average (WA)
CRC
beta=0.15, T=4
2026.06
97.6
84.8
14.8
Selective Prediction
beta=0.15, T=4
2026.06
97.3
85.1
14.5
Isotonic Confidence
beta=0.15, T=4
2026.06
97.3
85
14.6
Learned Stopper
beta=0.15, T=4
2026.06
96.6
83.6
16.4
Calibrated Learned
beta=0.15, T=4
2026.06
95.9
85.4
14
Consensus
beta=0.15, T=4
2026.06
94.5
84.5
15.5
Oracle
beta=0.15, T=4
2026.06
86
100
0
Confidence Threshold
beta=0.15, T=4
2026.06
69.8
84.5
15.5
kNN no bias
beta=0.15, T=4, ablate...
2026.06
37.9
97.7
0.9
Adaptive Local k-NN Certification (Ours)
beta=0.15, T=4
2026.06
29.8
98
0.6
Final-round
beta=0.15, T=4, bound_...
2026.06
29.3
97.9
0.6
Feedback
Search any
task
Search any
task