Metraipolis › archive › dc_c75cfcaba651
Direct or Mediated? Task-Dependent Audio Information Routing in Large Audio Language Models — arXiv:2608.27026
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.27026
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T18:54:06.226Z
- SHA-256 of the archived text
a90f72ef1b569ee8558a31a9547dbf03f9cc8097664a7e0becb9fcbfce404493- Stable URL of this page
https://metraipolis.com/documento/dc_c75cfcaba651
How to cite
Direct or Mediated? Task-Dependent Audio Information Routing in Large Audio Language Models — arXiv:2608.27026 — arXiv (en), https://arxiv.org/abs/2608.27026 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c75cfcaba651 (Metraipolis archive dc_c75cfcaba651, sha256 a90f72ef1b569ee8558a31a9547dbf03f9cc8097664a7e0becb9fcbfce404493)
BibTeX
@misc{dcc75cfcaba651,
title = {Direct or Mediated? Task-Dependent Audio Information Routing in Large Audio Language Models — arXiv:2608.27026},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.27026},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c75cfcaba651; sha256 a90f72ef1b569ee8558a31a9547dbf03f9cc8097664a7e0becb9fcbfce404493},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27026. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large Audio Language Models (LALMs) have demonstrated strong performance across a wide range of audio understanding tasks. However, they are typically evaluated on single, coherent audio segments, leaving their behavior under less familiar input configurations underexplored. We study this issue through a controlled setting in which two audio segments are concatenated into a single input. Across multiple LALMs, we observe a striking task-dependent robustness gap: automatic speech recognition (ASR) remains comparatively stable, whereas audio question answering (AQA) degrades substantially. To investigate the mechanisms underlying this disparity, we analyze how audio information is routed through LALM decoders using layer-wise attention knockout. The results reveal distinct task-dependent pathways. ASR relies primarily on direct retrieval from audio tokens by answer tokens, whereas AQA depends more strongly on a mediated route in which audio information is first integrated into prompt tokens and subsequently accessed during generation. We further probe prompt-token representations under audio concatenation and find that task-relevant audio attributes remain readily decodable, particularly in middle and later decoder layers, even when AQA performance deteriorates sharply. This dissociation indicates that the failure cannot be explained by complete loss of audio information from the decoder states and is instead consistent with a downstream bottleneck in retrieving or utilizing prompt-mediated information during answer generation. Together, our findings reveal task-dependent audio information routing in LALMs and highlight information utilization as a potential limitation on their generalization. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-28 in cs.MM). Categorie arXiv: Sound (cs.SD) ; Multimedia (cs.MM). Pagina del lavoro: https://arxiv.org/abs/2608.27026 — 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 2026-10-17T18:31:14.487Z dalla pagina https://arxiv.org/catchup/cs.MM/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).