Shanghai AI Lab Study: SFT Generalizes Under Three Conditions

上海AI Lab新研究：SFT能泛化，只要满足这三个条件

Shanghai AI Lab, Shanghai Jiao Tong University, and USTC tested Long-CoT SFT on Qwen3-14B-Base and found that cross-domain performance recovered and improved after 8 epochs, with generalization conditioned on optimization depth, data quality and structure, and base-model capability.

Why it matters: HKR-H/K/R all pass: the SFT-generalization claim has a clear hook, Qwen3-14B-Base plus an 8-epoch finding, and direct relevance to fine-tuning teams. It lacks deployment impact or full benchmark detail, so it stays in the mid-featured band.