SmolR1-SFT Potential limitations: Endless repetition Mistakes in reasoning Prompt format: ChatML Trained using Hugging Face's Open-R1 framework.