List your company

Mathematical Reasoning on GSM8K (Accuracy, Friedman Rank, P(EDEN))

81.5AccuracyBeam Search
69.22872.41475.678.786May 10, 2026

Evaluation results

MethodLinks
Beam Searchbeam width=3, total number of branches explored=12002026.0581.52.250.78
EDENtotal number of branches explored=5982026.0580.51.250.75
Majority-nn=52026.0578.76.250.99
Best-of-nn=52026.0578.24.620.96
Diverse Beam Searchbeam width=3, total number of branches explored=12002026.0577.54.750.96
Greedy2026.0573.56.120.99
Top-p Sampling2026.0573.56.620.99
Min-p Sampling2026.0572.381
Top-k Sampling2026.0570.771
Top-H Sampling2026.0569.78.121