Metraipolis › archive › dc_d9a773003edb

SkillFormer: Skill-Decomposed Adaptation for Audio Language Models — arXiv:2610.07533

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.07533
License (as archived)
CC0-1.0
Archived
2026-10-07T11:42:27.076Z
SHA-256 of the archived text
341e93b51b4e92a52a37233a4da4f93d967baabe0faf60442f8c00bd787f5adc
Stable URL of this page
https://metraipolis.com/documento/dc_d9a773003edb

How to cite

SkillFormer: Skill-Decomposed Adaptation for Audio Language Models — arXiv:2610.07533 — arXiv (en), https://arxiv.org/abs/2610.07533 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d9a773003edb (Metraipolis archive dc_d9a773003edb, sha256 341e93b51b4e92a52a37233a4da4f93d967baabe0faf60442f8c00bd787f5adc)
BibTeX
@misc{dcd9a773003edb,
  title = {SkillFormer: Skill-Decomposed Adaptation for Audio Language Models — arXiv:2610.07533},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.07533},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d9a773003edb; sha256 341e93b51b4e92a52a37233a4da4f93d967baabe0faf60442f8c00bd787f5adc},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07533. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Audio language models must handle dozens of distinct skills, from pitch comparison and speaker counting to musical tempo estimation and emotion recognition. Joint training on all skills at once causes interference: gains on one skill often come at the cost of another. We propose \textbf{SkillFormer}, which decomposes audio understanding into skill-specific low-rank adapters and composes them at inference time through a learned router. The router examines the question to decide which adapters to activate and how much weight each should carry, so that a pitch query engages different parameters than a genre classification query. An alternating training schedule updates each adapter on its own skill cluster before jointly calibrating the router, preventing the gradient conflicts that arise in standard multi-task optimization. SkillFormer adds fewer than 4\% of the base model's parameters and requires no changes to the audio encoder or language backbone. Evaluated on three architecturally distinct models across MMSU, MMAU-Pro, and MMAR, it raises the average accuracy by 2.5 to 4.1 points, with balanced gains across perception, reasoning, and semantic subcategories.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Sound (cs.SD) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.07533 — 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 2026-10-07T11:19:24.897Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).