Metraipolis › archive › dc_cdfdb0ed6f04
SR-OPSD: Self-Referenced On-Policy Self-Distillation — arXiv:2608.09745
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.09745
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T09:02:51.673Z
- SHA-256 of the archived text
02ff1f19ac14f85aaddb27a1b7b9a94e8981fa4ebc7a49fa55228819692e9c98- Stable URL of this page
https://metraipolis.com/documento/dc_cdfdb0ed6f04
How to cite
SR-OPSD: Self-Referenced On-Policy Self-Distillation — arXiv:2608.09745 — arXiv (en), https://arxiv.org/abs/2608.09745 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cdfdb0ed6f04 (Metraipolis archive dc_cdfdb0ed6f04, sha256 02ff1f19ac14f85aaddb27a1b7b9a94e8981fa4ebc7a49fa55228819692e9c98)
BibTeX
@misc{dccdfdb0ed6f04,
title = {SR-OPSD: Self-Referenced On-Policy Self-Distillation — arXiv:2608.09745},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.09745},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cdfdb0ed6f04; sha256 02ff1f19ac14f85aaddb27a1b7b9a94e8981fa4ebc7a49fa55228819692e9c98},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.09745. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
On-policy self-distillation (OPSD) converts feedback into dense token-level supervision on student-generated trajectories, complementing reinforcement learning with sparse outcome rewards. Its self-teacher, derived from the student's current or exponentially averaged parameters and conditioned on additional context, evolves alongside the student and its rollout context distribution. The benefit of modifying this moving target depends on how target--student probability mismatches translate into updates. We propose \emph{Self-Referenced On-Policy Self-Distillation (SR-OPSD)}, which constructs a normalized geometric target from the self-teacher and a frozen initial policy, then minimizes the forward Rényi divergence from this target to the student. The interpolation coefficient controls the self-teacher's contribution, while the Rényi order controls the power weighting of target-to-student probability ratios in the gradient. For fixed contexts and target components, we establish a conditional variational characterization and derive the exact token-logit gradient, revealing how anchoring and projection jointly shape the effective update target. Experiments across scientific reasoning, tool use, mathematical reasoning, and code generation demonstrate strong performance across multiple model families and scales. Ablations further show that reference anchoring can improve or degrade performance depending on the projection objective, supporting the joint design of target construction and projection geometry.
Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-11 in stat.ML).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2608.09745 — 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 2026-10-19T09:01:32.161Z dalla pagina https://arxiv.org/catchup/stat.ML/2026-08-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).