Metraipolis › archive › dc_c790cd25317d

Unsupervised Long-Tailed Adaptation of Vision-Language Models — arXiv:2610.07903

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.07903
License (as archived): CC0-1.0
Archived: 2026-10-07T13:07:21.677Z
SHA-256 of the archived text: aea9b806f7c4cfd671e006e4ca11e6f4e8910fe46bf14f3432961b4bb1eeb52a
Stable URL of this page: https://metraipolis.com/documento/dc_c790cd25317d

How to cite

Unsupervised Long-Tailed Adaptation of Vision-Language Models — arXiv:2610.07903 — arXiv (en), https://arxiv.org/abs/2610.07903 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c790cd25317d (Metraipolis archive dc_c790cd25317d, sha256 aea9b806f7c4cfd671e006e4ca11e6f4e8910fe46bf14f3432961b4bb1eeb52a)

BibTeX

@misc{dcc790cd25317d,
  title = {Unsupervised Long-Tailed Adaptation of Vision-Language Models — arXiv:2610.07903},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.07903},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c790cd25317d; sha256 aea9b806f7c4cfd671e006e4ca11e6f4e8910fe46bf14f3432961b4bb1eeb52a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07903. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Adapting vision-language models to downstream tasks has achieved remarkable success by leveraging pseudo-labels generated from unlabeled data. Existing methods typically assume a uniform unlabeled data distribution, and thus the resulting pseudo-label distribution is likewise uniform. However, real-world data distributions are often long-tailed. To tackle this, we formalize a new scenario termed Unsupervised Long-Tailed Adaptation (ULTA). Under this scenario, existing methods exhibit a contrasting phenomenon: head-class performance drops sharply, which is distinct from supervised long-tailed learning where tail classes suffer the most. In particular, we uncover that the distributional mismatch not only erodes head-class boundaries, but also pushes head samples into confusable classes, reinforcing the model's inherent bias. To address these issues, we propose a novel model called Margin-Aware Refinement with Structural alignment (MARS). Specifically, we mitigate head-class boundary erosion via Boundary-Preserving Alignment, which takes the zero-shot VLM as a fixed visual reference to suppress probability increases that lack visual support in the training targets. Building upon this, we introduce Margin-aware Self-Refinement, which employs a dynamic adjustment strategy to refine tail and confusable classes while preventing prediction bias. Extensive experiments on nine benchmark datasets demonstrate that MARS outperforms state-of-the-art methods, achieving an average accuracy improvement of 4.71 percentage points.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.07903 — 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 2026-10-07T12:37:32.925Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).