Authors

Xincheng Wei

Publications 2

Preprint Open access

Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation

Xincheng Wei, Yifan Ding, Yoshua Li et al. · 2026

On-policy self-distillation (OPSD) trains mathematical reasoning models using a privileged teacher that sees a reference solution and supervises student-sampled prefixes. Standard OPSD uses one fixed parameter setting at every state, but nearby settings may offer additional supervision. We find that local parameter per …

Preprint Open access

Do Coding Agents Reuse Existing Code or Reinvent the Wheel?

Dongsheng Ma, Sizhe Wang, Xinyi Huang et al. · 2026

Coding agents are increasingly deployed for iterative development on real repositories, yet existing evaluation barely answers a basic question: \emph{do coding agents reuse existing code or reinvent the wheel?} The question matters: every duplicated implementation is a fix applied twice and agents produce code far fas …

Co-authors

Yifan Ding Yoshua Li Yuquan Lu Ziheng Li Yi Lu Dongsheng Ma Rongxiang Weng Xunliang Cai Sizhe Wang Xinyi Huang Zhengren Wang Yuhan Wang