Metraipolis › archive › dc_4acf457e8a53
I Don't Miss You, but I Do: Self-Explanation Faithfulness of Modality Missingness in Vision-Language Models — arXiv:2609.07596
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.07596
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T12:57:18.054Z
- SHA-256 of the archived text
da6ed8e91df10c306d0fdd6c423f312cfdb0e92c722eb74581155ad36a8f9bad- Stable URL of this page
https://metraipolis.com/documento/dc_4acf457e8a53
How to cite
I Don't Miss You, but I Do: Self-Explanation Faithfulness of Modality Missingness in Vision-Language Models — arXiv:2609.07596 — arXiv (en), https://arxiv.org/abs/2609.07596 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4acf457e8a53 (Metraipolis archive dc_4acf457e8a53, sha256 da6ed8e91df10c306d0fdd6c423f312cfdb0e92c722eb74581155ad36a8f9bad)
BibTeX
@misc{dc4acf457e8a53,
title = {I Don't Miss You, but I Do: Self-Explanation Faithfulness of Modality Missingness in Vision-Language Models — arXiv:2609.07596},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.07596},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4acf457e8a53; sha256 da6ed8e91df10c306d0fdd6c423f312cfdb0e92c722eb74581155ad36a8f9bad},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.07596. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Vision-language models (VLMs) are increasingly used in settings where some input modalities may be unavailable, yet we know little about whether they can faithfully explain how such missing information affects their own predictions. We introduce an interventional protocol for evaluating self-explanations of modality dynamics: models state what each modality alone would support, whether restoring a missing modality would change their answer, and whether the available evidence is sufficient; we then execute the corresponding intervention and compare these claims with realized behavior. We evaluate ten VLMs spanning open-weight and proprietary models across four tasks covering mixed, redundant, and unique modality regimes. We find a systematic tendency to overstate the sufficiency of available evidence. Models substantially underestimate the effect of restoring missing modalities: executed change exceeds predicted change in 78 of 80 model-task-condition settings, with task-level median executed change rates reaching 70.1\% while median predicted rates remain at most 9.6\%. Insufficiency claims have low recall, leaving many cases in which behavior changes despite a stated claim of sufficiency. Retrospective self-explanations show the same tendency, over-crediting single-input sufficiency in mixed regimes and interchangeability in redundant ones. Together, these results show that VLMs systematically mischaracterize how their predictions depend on available and missing evidence, motivating executable interventions as a behavioral test of multimodal self-explanations. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CL). Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.07596 — 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 2026-10-16T12:30:25.313Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).