# Which Tokens Should SFT Actually Learn? A Token-Trimming Perspective on Mathematical Reasoning

> *Machine translation (AI-generated content)*

**Published:** September 10, 2026
**Author:** Yaning Jia, Chunhui Zhang, Wenxuan Xu, Xingjian Diao, Xiaoyuan Wang, Soroush Vosoughi
**Source:** arXiv — cs.AI
**Language:** en-US
**HTML:** https://aiqazaqstan.com/en/articles/which-tokens-should-sft-actually-learn-a-token-trimming-perspective-on-m

arXiv:2609.09707v1 Announce Type: new Abstract: Supervised fine-tuning (SFT) applies a uniform cross-entropy loss to all target tokens, even though different tokens provide unequal learning signals for mathematical reasoning. This uniform treatment can over-sharpen already…

> From the original source: **arXiv — cs.AI**
> [Read full article at arXiv — cs.AI](https://arxiv.org/abs/2609.09707)

**Tags:** #arXiv #Meta AI

---

*AI Qazaqstan — non-commercial informational portal about AI in Kazakhstan. Machine translations and summaries are AI-generated.*