List your company

Multilingual Mathematical Reasoning on MGSM 18 languages

72.5AccuracySP3F-7B
20.13633.730547.32560.9195Jan 26, 2026

Evaluation results

MethodLinks
SP3F-7BTraining Stage=Full Pipeline2026.0172.599.38
Qwen2.5-7B-InstructTraining Stage=Instruct2026.0166.3698.38
Qwen2.5-7B-Instruct + Translate TestTraining Stage=Instruct, Evaluation Protocol=Translate Test2026.0166.1595.81
Qwen2.5-7B + RLVRTraining Stage=SFT + RLVR2026.0165.3499.75
Qwen2.5-7B + SFTTraining Stage=SFT2026.0133.6691.37
Qwen2.5-7BTraining Stage=Base2026.0122.1590.67