Metraipolis › archive › dc_e71e20ce45e6
Why Vision Fails as a Universal Bridge: Rectifying Modality Asynchrony in Multilingual MLLMs — arXiv:2608.15085
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.15085
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T18:56:43.112Z
- SHA-256 of the archived text
dbbd678dbdb9f250c4461ba5f7d3165ecad207dd9771b7011d1274b92cb82fbb- Stable URL of this page
https://metraipolis.com/documento/dc_e71e20ce45e6
How to cite
Why Vision Fails as a Universal Bridge: Rectifying Modality Asynchrony in Multilingual MLLMs — arXiv:2608.15085 — arXiv (en), https://arxiv.org/abs/2608.15085 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e71e20ce45e6 (Metraipolis archive dc_e71e20ce45e6, sha256 dbbd678dbdb9f250c4461ba5f7d3165ecad207dd9771b7011d1274b92cb82fbb)
BibTeX
@misc{dce71e20ce45e6,
title = {Why Vision Fails as a Universal Bridge: Rectifying Modality Asynchrony in Multilingual MLLMs — arXiv:2608.15085},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.15085},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e71e20ce45e6; sha256 dbbd678dbdb9f250c4461ba5f7d3165ecad207dd9771b7011d1274b92cb82fbb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15085. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models (MLLMs) exhibit substantial performance degradation in non-English visual reasoning, despite the strong multilingual competence of their text-only backbones. While mechanistic evidence from text-only models suggests that non-English inputs are routed through an English-centric latent space, the multimodal implications of this phenomenon remain unexplored. Through rigorous mechanistic analysis, we identify the \textbf{Ghost Anchor} phenomenon: a temporal modality asynchrony where linguistic translation to the English semantic manifold completes in early layers, while visual semanticization remains immature. Consequently, visual signals are physically present yet functionally invisible during the early alignment window. To rectify this, we propose \textbf{ANCHOR}, a training framework employing Proactive Visual Anchoring (PVA) to accelerate early visual semantic emergence, ensuring visual representations proactively guide linguistic translation. Mechanistic interventions confirm that ANCHOR successfully restores the causal influence of visual signals during early translation. Furthermore, extensive experiments on XMMMU, MaXM, and CVQA demonstrate that ANCHOR consistently outperforms standard baselines, achieving robust visual reasoning across both fine-tuned and zero-shot languages.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.15085 — 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 2026-10-18T18:06:57.844Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).