Metraipolis › archive › dc_535dca167c29

Divergence-Routed On-Policy Distillation for Language Models — arXiv:2610.19572

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.19572
License (as archived)
CC0-1.0
Archived
2026-10-15T20:01:49.129Z
SHA-256 of the archived text
7c9ffd91cb5349606683f897181100fa24f2026301edf0b7e685ac91edc6f2e5
Stable URL of this page
https://metraipolis.com/documento/dc_535dca167c29

How to cite

Divergence-Routed On-Policy Distillation for Language Models — arXiv:2610.19572 — arXiv (en), https://arxiv.org/abs/2610.19572 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_535dca167c29 (Metraipolis archive dc_535dca167c29, sha256 7c9ffd91cb5349606683f897181100fa24f2026301edf0b7e685ac91edc6f2e5)
BibTeX
@misc{dc535dca167c29,
  title = {Divergence-Routed On-Policy Distillation for Language Models — arXiv:2610.19572},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.19572},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_535dca167c29; sha256 7c9ffd91cb5349606683f897181100fa24f2026301edf0b7e685ac91edc6f2e5},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.19572. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

On-policy distillation (OPD) transfers knowledge from a teacher model to a student along student-generated trajectories, typically by minimizing the reverse Kullback-Leibler (RKL) divergence. However, RKL can provide weak corrective signals for teacher-supported tokens to which the student assigns very low probability. This raises two questions: which positions are most consequential for the behavioral gap between the student and teacher, and what additional supervision is needed at these positions? Through cross-sampling, we find that divergence-based routing more effectively pinpoints positions contributing to the teacher-student performance gap than entropy-based routing. Further analysis reveals teacher-mode under-coverage at high Jensen-Shannon (JS) divergence positions, where standard OPD increases the large teacher-student probability gaps. Motivated by these findings, we propose Divergence-Routed On-Policy Distillation (DrOPD). DrOPD retains RKL supervision at all positions and selectively adds forward Kullback-Leibler (FKL) divergence supervision at positions with high JS divergence. Experiments across six math benchmarks show that DrOPD improves over standard OPD and entropy-based routing. DrOPD also reduces severe under-coverage cases by 37.9 percent relative to OPD, supporting targeted FKL as an effective complement to OPD.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.19572 — 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 2026-10-15T19:35:48.249Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).