Paper recorded by Signals 4 on 2026-09-24 in cs.LG. Abstract reproduced from arXiv; link to the original below.
Published 2026-09-24 on arXiv · recorded by Signals 4 on 2026-09-25
Category: cs.LG · 机器学习 · first seen 2026-09-25
Layer-wise visual-text similarity in Multimodal Large Language Models (MLLMs) is widely interpreted as evidence that the language model progressively integrates visual content into a shared representation space. This reading rests on the assumption that scalar alignment scores reflect content-level cross-modal interaction. To test this assumption, we apply controlled interventions to the visual st