Qwen2.5-Math-15B GRPO ACEMath Fine-tune (Jongbin-kr)

A community fine-tune of Alibaba's Qwen2.5-Math-15B base model, applying GRPO-style reinforcement learning post-training on the ACEMath dataset to improve step-by-step mathematical reasoning. It is not a standalone product or organization, but a research checkpoint uploaded to Hugging Face by individual user Jongbin-kr. Source: supabase.