Metraipolis › archive › dc_3af7eae7cd83
Temporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision — arXiv:2609.04203
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04203
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T02:56:38.540Z
- SHA-256 of the archived text
acb37b8ab021483d0a1ea42289c37f29ef34f835f35d2bdf784f851b31fccf36- Stable URL of this page
https://metraipolis.com/documento/dc_3af7eae7cd83
How to cite
Temporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision — arXiv:2609.04203 — arXiv (en), https://arxiv.org/abs/2609.04203 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3af7eae7cd83 (Metraipolis archive dc_3af7eae7cd83, sha256 acb37b8ab021483d0a1ea42289c37f29ef34f835f35d2bdf784f851b31fccf36)
BibTeX
@misc{dc3af7eae7cd83,
title = {Temporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision — arXiv:2609.04203},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04203},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3af7eae7cd83; sha256 acb37b8ab021483d0a1ea42289c37f29ef34f835f35d2bdf784f851b31fccf36},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04203. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We introduce S$^3$T (Self-Supervised Self-Distillation over Time), which, to the best of our knowledge, is the first fully self-contained framework for continuous video state tracking. Our method treats temporal sampling density as privileged information, based on the hypothesis that a denser view of the same clip recovers the running state more accurately. This view serves as the teacher, while a sparse-view student with the same weights learns to match its next-token distribution. The model generates its own target, so training requires no labels, separate teacher, or reward signal, and adds no inference cost. On LLaVA-OneVision-2-8B, S$^3$T improves VSTAT accuracy by $+1.74$ as a single model, $+2.38$ with souping, and $+2.70$ with additional vision-encoder adaptation, while prior self-evolving methods leave state tracking largely unchanged. The capability learned from unlabeled synthetic clips transfers to real videos, improving performance by $+7.95$ on VSTAT-YouTube state-tracking questions and $+4.50$ on MVBench Action Count. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-04 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.04203 — 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 2026-10-17T02:27:19.021Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).