List your company

Mathematical Problem Solving on Minerva (Pass@1)

33.32Pass@1VHG (Soft)
23.356825.943428.5331.1166May 7, 2026

Evaluation results

MethodLinks
VHG (Soft)2026.0533.32
Vanilla-GRPOtraining_variant=w. SFT data2026.0532.63
Vanilla-GRPO2026.0531.92
R-ZeroIteration=22026.0528.52
R-ZeroIteration=32026.0528.12
Qwen3-4B-Base2026.0527.78
R-ZeroIteration=12026.0523.74