Replication of fine-tuning the Qwen2.5 family of models on the S1 and S1.1 datasets, as described in the S1 work (https://arxiv.org/abs/2501.19393)