Mathematical Reasoning on OlympiadBench, AIME24, AIME25, and Minerva (Average)
27.7Average ScoreQwen3-4B-GRLO+RLVR
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-4B-GRLO+RLVRBackbone=Qwen3-4B, Training=GRLO+RLVR2026.05 | 27.7 | |
| Qwen3-4B-GRLOBackbone=Qwen3-4B, Training=GRLO2026.05 | 24.9 | |
| General-Reasoner-4BBackbone=General-Reasoner-4B2026.05 | 24.4 | |
| Qwen3-4B-BaseBackbone=Qwen3-4B2026.05 | 19.5 |