@misc{indiciaee7f6af8649ed,
  title = {Rethinking Supervised Fine-Tuning: Emphasizing Key Answer Tokens for Improved LLM Accuracy},
  author = {Xiaofeng Shi and Qian Kou and Yuduo Li and Hua Zhou},
  year = {2025},
  url = {https://arxiv.org/abs/2512.21017},
  note = {Source identifier: 2512.21017}
}