Metraipolis › archive › dc_44d0877642af
Mid-Training Language Models on Raw Video — arXiv:2610.11019
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.11019
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T05:56:16.373Z
- SHA-256 of the archived text
6598209ef72fadc0572fff31b885bbd5a2885c3e763aab52b705025f1f7c4888- Stable URL of this page
https://metraipolis.com/documento/dc_44d0877642af
How to cite
Mid-Training Language Models on Raw Video — arXiv:2610.11019 — arXiv (en), https://arxiv.org/abs/2610.11019 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_44d0877642af (Metraipolis archive dc_44d0877642af, sha256 6598209ef72fadc0572fff31b885bbd5a2885c3e763aab52b705025f1f7c4888)
BibTeX
@misc{dc44d0877642af,
title = {Mid-Training Language Models on Raw Video — arXiv:2610.11019},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.11019},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_44d0877642af; sha256 6598209ef72fadc0572fff31b885bbd5a2885c3e763aab52b705025f1f7c4888},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11019. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models learn mostly from paired image-text data or annotated video, and raw web video is rarely used to further train an existing language model. We study whether raw video, with no captions and no text loss, can serve as mid-training data for a pretrained language model. Frames are encoded into continuous visual tokens, and the language model learns to predict the next visual token. We mid-train Qwen3-1.7B on raw clips from YT-Temporal-1B and then apply the same image-text instruction tuning to it and to the model without mid-training, so that the two differ only in mid-training. The mid-trained model scores 2.9 points higher on average across four video benchmarks and 5.1 points higher across ten image benchmarks, spanning perception, document, and chart tasks. Text performance is preserved even though mid-training includes no text, with an average of 48.9 across 14 text benchmarks compared with 48.0 for the model without mid-training. Analyses across training show that the image and video gains emerge within 30% of training and plateau thereafter, varying by less than 0.5 points. Predicting captions fails to outperform next-visual-token prediction, demonstrating that video mid-training can remain purely self-supervised without the computational overhead or labeling noise of automated captioning. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.11019 — 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 2026-10-09T05:47:01.487Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).