List your company

Mathematical Reasoning on AIME25 (Accuracy)

76.3AccuracyLUSPO
0.69220.32139.9559.579Feb 5, 2026

Evaluation results

MethodLinks
LUSPOBase model=Qwen3-30B-A3B-Instruct2026.0276.3
GSPOBase model=Qwen3-30B-A3B-Instruct2026.0259.2
w/o RLVRBase model=Qwen3-30B-A3B-Instruct2026.0257.2
LUSPOBase model=Qwen2.5-7B-Base2026.0213.9
GSPOBase model=Qwen2.5-7B-Base2026.0211.2
w/o RLVRBase model=Qwen2.5-7B-Base2026.023.6