Metraipolis › archive › dc_3b57df8cb8a7
On the Few-Shot Adaptation of Vision-Language Models: A Few Gradient Signs Are All You Need — arXiv:2611.08800
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.08800
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-09T22:41:16.669Z
- SHA-256 of the archived text
66e26db780776e77788a67c246e250e68ba8b8255702f261b4a808d3b5ef028a- Stable URL of this page
https://metraipolis.com/documento/dc_3b57df8cb8a7
How to cite
On the Few-Shot Adaptation of Vision-Language Models: A Few Gradient Signs Are All You Need — arXiv:2611.08800 — arXiv (en), https://arxiv.org/abs/2611.08800 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3b57df8cb8a7 (Metraipolis archive dc_3b57df8cb8a7, sha256 66e26db780776e77788a67c246e250e68ba8b8255702f261b4a808d3b5ef028a)
BibTeX
@misc{dc3b57df8cb8a7,
title = {On the Few-Shot Adaptation of Vision-Language Models: A Few Gradient Signs Are All You Need — arXiv:2611.08800},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.08800},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3b57df8cb8a7; sha256 66e26db780776e77788a67c246e250e68ba8b8255702f261b4a808d3b5ef028a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.08800. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
The limited supervision available for the few-shot adaptation of vision-language models makes gradient estimation particularly noisy. We find that the gradient magnitudes are severely miscalibrated, while the signs of a small top-magnitude subset remain informative. This motivates SignSparseSGD and DynSignSparseSGD, two state-free optimizers that use gradient magnitudes only to select where to update and fixed-magnitude signs to decide how to update. The static variant fixes the support after the first minibatch. The dynamic variant refreshes it at every step. Across 11 datasets, three CLIP backbones, and five shot levels, the proposed methods are competitive with or outperform parameter-efficient and optimization-based baselines on average, while exhibiting markedly lower sensitivity to long training schedules. They introduce no task-specific modules, maintain no first- or second-moment optimizer states, and use one shared configuration across datasets, backbones, and shot levels. Our code is included in the supplementary materials and will be released upon publication. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2611.08800 — 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 2026-11-09T22:33:47.142Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).