We have released two preprints: one that comprehensively investigates the effects of supervised fine-tuning on LLMs arXiv, and another that examines the impact of internal language representations arXiv.