List your company

Math Reasoning on mGSM v2

78Accuracy (Seen)GRPO
21.486436.158250.8365.5018May 25, 2026

Evaluation results

MethodLinks
GRPOBackbone=Gemma3-4b-it2026.057859.271.16
LRPOBackbone=Gemma3-4b-it2026.0577.4962.171.89
VanillaBackbone=Gemma3-4b-it2026.0577.2659.270.69
MAPOBackbone=Gemma3-4b-it2026.0576.6760.270.68
DPOBackbone=Gemma3-4b-it2026.0576.3458.970
LIDRBackbone=Gemma3-4b-it2026.0576.2358.969.93
MPOBackbone=Gemma3-4b-it2026.0575.9459.369.89
LRPOBackbone=Qwen2.5-1.5b-it2026.0552.5113.338.25
GRPOBackbone=Qwen2.5-1.5b-it2026.0543.4312.932.33
DPOBackbone=Qwen2.5-1.5b-it2026.0537.3314.6427.02
MAPOBackbone=Qwen2.5-1.5b-it2026.0536.812.2425.64
VanillaBackbone=Qwen2.5-1.5b-it2026.0535.7311.8424.87
MPOBackbone=Qwen2.5-1.5b-it2026.0535.6712.3225.05
LIDRBackbone=Qwen2.5-1.5b-it2026.0534.210.8823.6
DPOBackbone=LLaMA3.2-1b-it2026.052612.821.2
LRPOBackbone=LLaMA3.2-1b-it2026.0525.3112.220.55
MAPOBackbone=LLaMA3.2-1b-it2026.0525.0911.220.04
GRPOBackbone=LLaMA3.2-1b-it2026.0524.9114.821.24
MPOBackbone=LLaMA3.2-1b-it2026.0524.51918.87
LIDRBackbone=LLaMA3.2-1b-it2026.0523.8910.719.09
VanillaBackbone=LLaMA3.2-1b-it2026.0523.668.718.22