Metraipolis › archive › dc_aed3df959afe
Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning — arXiv:2609.32667
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.32667
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:55:51.248Z
- SHA-256 of the archived text
1ec75bf98d1e62f70f68f751f4a5e6d6cf2bf807ab3a50d95d7daf7f92013503- Stable URL of this page
https://metraipolis.com/documento/dc_aed3df959afe
How to cite
Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning — arXiv:2609.32667 — arXiv (en), https://arxiv.org/abs/2609.32667 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_aed3df959afe (Metraipolis archive dc_aed3df959afe, sha256 1ec75bf98d1e62f70f68f751f4a5e6d6cf2bf807ab3a50d95d7daf7f92013503)
BibTeX
@misc{dcaed3df959afe,
title = {Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning — arXiv:2609.32667},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.32667},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_aed3df959afe; sha256 1ec75bf98d1e62f70f68f751f4a5e6d6cf2bf807ab3a50d95d7daf7f92013503},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32667. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
On-policy self-distillation (OPSD) trains a question-only student with token-level feedback from a teacher given training-only privileged information (PI). OPSD therefore provides dense, on-policy supervision, and is free of a larger external teacher, but its effectiveness rests on how PI is designed and utilized. Our preliminary diagnostics suggest a significant gap between teacher utility and student learnability, where a small fraction of high-disagreement tokens dominate the distillation signal, and short teacher continuations at these positions further expose more explicit PI leakage than transferable correction cues, indicating a strong intent on injecting PI-conditioned shortcuts. We propose Adaptive On-Policy Self-Distillation (AOPSD), which adapts what information the teacher receives and how strongly its feedback influences learning. AOPSD encodes each solution as a reasoning DAG, orders problems by the student's evolving capability, and reveals only the affordable subgraph and its next frontier as PI. For high-disagreement tokens, AOPSD utilizes short teacher continuations as probes to encourage useful guidance while mitigating PI-conditioned shortcuts among teacher supervisions. On HMMT25, AIME24, AIME25, and BRUMo25, AOPSD achieves 72.5% Pass@8, which is 6.7 percentage points above OPSD and 4.2 above the strongest competing baseline while reducing 15 percentage points of training time at lower cost. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.32667 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T20:29:00.911Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).