List your company

Mathematical Reasoning on Minerva Math (Score)

42.6ScoreQwen3-4B-GRLO
19.7225.6631.637.54May 14, 2026

Evaluation results

MethodLinks
Qwen3-4B-GRLOBackbone=Qwen3-4B, Training=GRLO2026.0542.6
Qwen3-4B-GRLO+RLVRBackbone=Qwen3-4B, Training=GRLO+RLVR2026.0541.5
General-Reasoner-4BBackbone=General-Reasoner-4B2026.0525.7
Qwen3-4B-BaseBackbone=Qwen3-4B2026.0520.6