发现微调时中间层对指令理解最关键，只优化这些层就能更好对齐模型。
A Layer-wise Analysis of Supervised Fine-Tuning
- 分析不同层级在微调中的变化，发现中间层最稳定、末层最敏感。
- 新方法仅更新中间层，比标准LoRA在GSM8K上提升10.2%。
- 适合追求高效对齐的开发者，尤其关注参数效率的场景。
尽管对齐至关重要，监督微调（SFT）存在灾难性遗忘风险，且指令遵循能力的层级演化机制仍不明确。我们通过信息论、几何与优化指标，对1B-32B规模模型进行了全面分析。实验揭示显著的深度依赖模式：中间层（20%-80%）保持稳定，而最终层高度敏感。基于此，我们提出中段块高效微调（Mid-Block Efficient Tuning），仅选择性更新关键中间层。实证表明，该方法在GSM8K（OLMo2-7B）上优于标准LoRA达10.2%，且参数开销更低，证明有效对齐具有架构局部性而非全局分布性。代码已公开于https://anonymous.4open.science/r/base_sft。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。