Metraipolis › archive › dc_2f9374a20cf2
When Does the Concept of "Dog" Emerge in an Audio LLM? — arXiv:2609.33458
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.33458
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:56:49.209Z
- SHA-256 of the archived text
6c21505565923f2d63a709aed835fe7c175006b98860ffa2f05b0f7c40e58df6- Stable URL of this page
https://metraipolis.com/documento/dc_2f9374a20cf2
How to cite
When Does the Concept of "Dog" Emerge in an Audio LLM? — arXiv:2609.33458 — arXiv (en), https://arxiv.org/abs/2609.33458 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2f9374a20cf2 (Metraipolis archive dc_2f9374a20cf2, sha256 6c21505565923f2d63a709aed835fe7c175006b98860ffa2f05b0f7c40e58df6)
BibTeX
@misc{dc2f9374a20cf2,
title = {When Does the Concept of "Dog" Emerge in an Audio LLM? — arXiv:2609.33458},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.33458},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2f9374a20cf2; sha256 6c21505565923f2d63a709aed835fe7c175006b98860ffa2f05b0f7c40e58df6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33458. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models answer audio questions, but how they represent auditory semantics and use them in decisions remains unclear, limiting our understanding of response formation. We study dog barking in Qwen2.5-Omni-7B using Jacobian lens (J-lens) readout and directional interventions. We define the dog direction as a J-lens-derived hidden-state vector associated with dog; adding or removing its component modulates dog-related information. We find this information decodable without dog/bark prompt cues or animal-identification requirements. Directional interventions change response tendencies and some final answers, with effects concentrated in late-layer states immediately before generation across species classification, vocalization classification, and sound description. The dog direction shows no comparable advantage over controls in animal/other classification. These results provide causal-intervention evidence that the dog direction affects output scores in a task-dependent manner, most consistently at L22 and L24 immediately before generation. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.33458 — 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 2026-10-07T20:29:01.266Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).