pipette
ENEnglish

Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning

Yuxuan Wang, Guangxiang Zhao, Xinyu Tan, Yahui Han, Xinxiang Yin, Zhonghai Wu, Xiangzheng Zhang, Lin Sun, Tong Yang

Preprint

En palabras de los autores

Supervised fine-tuning (SFT) changes logit margins among non-target tokens at teacher-forced prefixes. Matching a local reference distribution keeps those probabilities and also penalizes margin expansions. We ask whether a soft auxiliary can preserve selected local non-target margin relations without freezing that distribution. We introduce Selective Rank-Preserving SFT (SRP-SFT), a soft margin target: a hinge on reference-scaled lower bounds for pairs from the reference's top- non-target tokens, applied at fixed teacher-forced prefixes. The hinge stops once the requested bound is met. We derive the KL cost of violating a margin, characterize the feasible region and its conditional utility, and give a finite-step condition for reducing aggregate violations. In one-epoch experiments on Qwen3-4B-Base and UltraFeedback, this target raises MATH-500 and code-generation scores over CE and LP-SFT at all three training seeds. At seed 1234, MATH-500 pass@1 rises from 49.56% to 54.56% and the four-metric code mean rises from 65.91% to 69.50% relative to LP-SFT. Ranking controls, held-out cache diagnostics, sensitivity sweeps, and six additional backbones show where these score changes appear.

Resultado principalLimitación que admiten los autores

Apareció: jueves, 15 de octubre. arXiv. Preprint, todavía sin revisión por pares.