While critical for alignment, Supervised Fine-Tuning (SFT) incurs the risk of catastrophic forgetting, yet the layer-wise emergence of instruction-following capabilities remains elusive. We investigate this mechanism via a comprehensive analysis utilizing information-theoretic, geometric, and optimization metrics across model scales (1B-32B). Our experiments reveal a distinct depth-dependent pattern: middle layers (20\%-80\%) are stable, whereas final layers exhibit high sensitivity. Leveraging this insight, we propose Mid-Block Efficient Tuning, which selectively updates these critical intermediate layers. Empirically, our method outperforms standard LoRA up to 10.2\% on GSM8K (OLMo2-7B) with reduced parameter overhead, demonstrating that effective alignment is architecturally localized rather than distributed. The code is publicly available at https://anonymous.4open.science/r/base_sft.
@article{arxiv.2604.11838,
title = {A Layer-wise Analysis of Supervised Fine-Tuning},
author = {Qinghua Zhao and Xueling Gong and Xinyu Chen and Zhongfeng Kang and Xinlu Li},
journal= {arXiv preprint arXiv:2604.11838},
year = {2026}
}