Metraipolis › archive › dc_61ebaa2684bd
Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation — arXiv:2608.14655
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.14655
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T19:39:45.928Z
- SHA-256 of the archived text
b3c151f00ce092716fb87e3133a8c5dae67c90c9fff91a3e57012d2d6c50f6bb- Stable URL of this page
https://metraipolis.com/documento/dc_61ebaa2684bd
How to cite
Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation — arXiv:2608.14655 — arXiv (en), https://arxiv.org/abs/2608.14655 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_61ebaa2684bd (Metraipolis archive dc_61ebaa2684bd, sha256 b3c151f00ce092716fb87e3133a8c5dae67c90c9fff91a3e57012d2d6c50f6bb)
BibTeX
@misc{dc61ebaa2684bd,
title = {Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation — arXiv:2608.14655},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.14655},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_61ebaa2684bd; sha256 b3c151f00ce092716fb87e3133a8c5dae67c90c9fff91a3e57012d2d6c50f6bb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.14655. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Omni-Large Language Models (Omni-LLMs) power complex multi-modal reasoning in applications like World Action Models and autonomous agents. However, their strong performance often masks a profound Perceptual-Decision Misalignment (PDM), where decisions remain unfaithful to multi-modal perceptions. To diagnose this, we formalize Causal Modality Sensitivity (CMS), operationalized via a dual-lens framework: Answer Retention Rate (ARR) at the macro behavioral level, and Logit Angular Discrepancy (LAD) to track microscopic distribution shifts. We also curate CausalMSBench, a diagnostic dataset isolating language priors. Benchmarking reveals that popular Omni-LLMs exhibit critically low CMS, showing negligible distribution shifts even when key modalities are removed. To rectify this, we propose Modality Subspace Activation (MSA), a training-free inference-time framework that uses Singular Value Decomposition (SVD) to estimate modal activation strengths. MSA dynamically balances modal projections in the last hidden state, effectively restoring CMS across benchmarks. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.LG). Categorie arXiv: Machine Learning (cs.LG) ; Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2608.14655 — 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 2026-10-18T19:02:44.031Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).