Reproduce “Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jiacheng Du, Weiwei Xie, Tianyi Du, Shaoxiong Guo, Qibing Ren, Jiaheng Zhang
AOPSD trains a question-only math student using a frozen copy of its base model as a teacher, selectively revealing solution-DAG steps and weighting large-disagreement feedback with a student-scored continuation. On Qwen3-4B-Instruct-2507, 725 LIMO-V2 problems, and four benchmarks, it reaches a 72.5% average Pass@8, versus 65.9% for OPSD; this result is limited to that setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.