Metraipolis › archive › dc_33786a4a5fcb
Latent On-Policy Self-Distillation — arXiv:2608.13040
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.13040
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T04:24:42.604Z
- SHA-256 of the archived text
498cb51733dc432658f4a47c801c197af454fd07823128f37a0661efd04f30e5- Stable URL of this page
https://metraipolis.com/documento/dc_33786a4a5fcb
How to cite
Latent On-Policy Self-Distillation — arXiv:2608.13040 — arXiv (en), https://arxiv.org/abs/2608.13040 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_33786a4a5fcb (Metraipolis archive dc_33786a4a5fcb, sha256 498cb51733dc432658f4a47c801c197af454fd07823128f37a0661efd04f30e5)
BibTeX
@misc{dc33786a4a5fcb,
title = {Latent On-Policy Self-Distillation — arXiv:2608.13040},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.13040},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_33786a4a5fcb; sha256 498cb51733dc432658f4a47c801c197af454fd07823128f37a0661efd04f30e5},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.13040. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Enabling agents to learn from experience and internalize it into their policy has become a central problem in self-evolving AI. On-policy self-distillation (OPSD) offers an effective pathway by using a privileged self-teacher to provide dense supervision on the student's own trajectories; however, existing methods still rely heavily on designer-specified privileged artifacts (e.g., answers, feedback, skills, or trajectories), limiting the end-to-end learnability and scalability required for continual self-improvement. In this work, we introduce Latent On-Policy Self-Distillation (LOPD), which, rather than proposing another hand-crafted OPSD variant with a newly prescribed form of privileged context, makes the teacher's privileged context itself learnable end-to-end from experience. Technically, LOPD retrieves relevant experiences and composes them into continuous latent tokens that condition a self-teacher, while the student generates trajectories from the task and interaction history and receives dense token-level supervision at every visited prefix. We further introduce a privileged-margin objective to stabilize and regulate the learning of latent context. Empirically, LOPD demonstrates (I) strong performance, outperforming RLVR and representative OPSD methods including OPSD, SDPO, and Skill-SD across both agentic tool use and code generation; and (II) high learning efficiency, surpassing GRPO and Skill-SD with less than 30% of their rollout budget. Ablation studies further provide direct evidence that making privileged context learnable is necessary for realizing these gains. Together, these results position LOPD as a step toward a more scalable and self-directed paradigm for agent evolution. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-14 in cs.CL). Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.13040 — 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 2026-10-19T04:17:42.413Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-14?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).